diff --git a/.github/workflows/build-docker.yml b/.github/workflows/build-docker.yml index d77958b..d2869ed 100644 --- a/.github/workflows/build-docker.yml +++ b/.github/workflows/build-docker.yml @@ -2,7 +2,7 @@ name: CD on: push: - branches: [ "master" ] + branches: [ "main" ] workflow_dispatch: # run manually jobs: diff --git a/.github/workflows/test-eynollah.yml b/.github/workflows/test-eynollah.yml index b27586c..042e508 100644 --- a/.github/workflows/test-eynollah.yml +++ b/.github/workflows/test-eynollah.yml @@ -27,7 +27,12 @@ jobs: - uses: actions/cache@v4 id: seg_model_cache with: - path: models_eynollah + path: models_layout_v0_5_0 + key: ${{ runner.os }}-models + - uses: actions/cache@v4 + id: ocr_model_cache + with: + path: models_ocr_v0_5_0 key: ${{ runner.os }}-models - uses: actions/cache@v4 id: bin_model_cache @@ -35,7 +40,7 @@ jobs: path: default-2021-03-09 key: ${{ runner.os }}-modelbin - name: Download models - if: steps.seg_model_cache.outputs.cache-hit != 'true' || steps.bin_model_cache.outputs.cache-hit != 'true' + if: steps.seg_model_cache.outputs.cache-hit != 'true' || steps.bin_model_cache.outputs.cache-hit != 'true' || steps.ocr_model_cache.outputs.cache-hit != true run: make models - name: Set up Python ${{ matrix.python-version }} uses: actions/setup-python@v5 diff --git a/.gitignore b/.gitignore index 5236dde..0d5d834 100644 --- a/.gitignore +++ b/.gitignore @@ -5,3 +5,4 @@ models_eynollah* output.html /build /dist +*.tif diff --git a/CHANGELOG.md b/CHANGELOG.md index ad86fe5..0ad9a09 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -5,9 +5,17 @@ Versioned according to [Semantic Versioning](http://semver.org/). ## Unreleased +## [0.5.0] - 2025-09-26 + Fixed: - * restoring the contour in the original image caused an error due to an empty tuple + * restoring the contour in the original image caused an error due to an empty tuple, #154 + +Added: + + * `eynollah machine-based-reading-order` CLI to run reading order detection, #175 + * `eynollah enhancement` CLI to run image enhancement, #175 + * Improved models for page extraction and reading order detection, #175 ## [0.4.0] - 2025-04-07 @@ -187,6 +195,8 @@ Fixed: Initial release +[0.5.0]: ../../compare/v0.5.0...v0.4.0 +[0.4.0]: ../../compare/v0.4.0...v0.3.1 [0.3.1]: ../../compare/v0.3.1...v0.3.0 [0.3.0]: ../../compare/v0.3.0...v0.2.0 [0.2.0]: ../../compare/v0.2.0...v0.1.0 diff --git a/Makefile b/Makefile index 73d4d34..a920615 100644 --- a/Makefile +++ b/Makefile @@ -9,12 +9,15 @@ DOCKER ?= docker #SEG_MODEL := https://qurator-data.de/eynollah/2021-04-25/models_eynollah.tar.gz #SEG_MODEL := https://qurator-data.de/eynollah/2022-04-05/models_eynollah_renamed.tar.gz -SEG_MODEL := https://qurator-data.de/eynollah/2022-04-05/models_eynollah.tar.gz +# SEG_MODEL := https://qurator-data.de/eynollah/2022-04-05/models_eynollah.tar.gz #SEG_MODEL := https://github.com/qurator-spk/eynollah/releases/download/v0.3.0/models_eynollah.tar.gz #SEG_MODEL := https://github.com/qurator-spk/eynollah/releases/download/v0.3.1/models_eynollah.tar.gz +SEG_MODEL := https://zenodo.org/records/17194824/files/models_layout_v0_5_0.tar.gz?download=1 BIN_MODEL := https://github.com/qurator-spk/sbb_binarization/releases/download/v0.0.11/saved_model_2021_03_09.zip +OCR_MODEL := https://zenodo.org/records/17194824/files/models_ocr_v0_5_0.tar.gz?download=1 + PYTEST_ARGS ?= -vv # BEGIN-EVAL makefile-parser --make-help Makefile @@ -28,7 +31,7 @@ help: @echo " install Install package with pip" @echo " install-dev Install editable with pip" @echo " deps-test Install test dependencies with pip" - @echo " models Download and extract models to $(CURDIR)/models_eynollah" + @echo " models Download and extract models to $(CURDIR)/models_layout_v0_5_0" @echo " smoke-test Run simple CLI check" @echo " ocrd-test Run OCR-D CLI check" @echo " test Run unit tests" @@ -44,14 +47,20 @@ help: # END-EVAL -# Download and extract models to $(PWD)/models_eynollah -models: models_eynollah default-2021-03-09 +# Download and extract models to $(PWD)/models_layout_v0_5_0 +models: models_layout_v0_5_0 models_ocr_v0_5_0 default-2021-03-09 -models_eynollah: models_eynollah.tar.gz - tar zxf models_eynollah.tar.gz +models_layout_v0_5_0: models_layout_v0_5_0.tar.gz + tar zxf models_layout_v0_5_0.tar.gz -models_eynollah.tar.gz: - wget $(SEG_MODEL) +models_layout_v0_5_0.tar.gz: + wget -O $@ $(SEG_MODEL) + +models_ocr_v0_5_0: models_ocr_v0_5_0.tar.gz + tar zxf models_ocr_v0_5_0.tar.gz + +models_ocr_v0_5_0.tar.gz: + wget -O $@ $(OCR_MODEL) default-2021-03-09: $(notdir $(BIN_MODEL)) unzip $(notdir $(BIN_MODEL)) @@ -73,20 +82,28 @@ install: install-dev: $(PIP) install -e .$(and $(EXTRAS),[$(EXTRAS)]) -deps-test: models_eynollah +deps-test: models_layout_v0_5_0 $(PIP) install -r requirements-test.txt smoke-test: TMPDIR != mktemp -d smoke-test: tests/resources/kant_aufklaerung_1784_0020.tif # layout analysis: - eynollah layout -i $< -o $(TMPDIR) -m $(CURDIR)/models_eynollah + eynollah layout -i $< -o $(TMPDIR) -m $(CURDIR)/models_layout_v0_5_0 fgrep -q http://schema.primaresearch.org/PAGE/gts/pagecontent/2019-07-15 $(TMPDIR)/$(basename $( Document Layout Analysis with Deep Learning and Heuristics + +> Document Layout Analysis, Binarization and OCR with Deep Learning and Heuristics [![PyPI Version](https://img.shields.io/pypi/v/eynollah)](https://pypi.org/project/eynollah/) [![GH Actions Test](https://github.com/qurator-spk/eynollah/actions/workflows/test-eynollah.yml/badge.svg)](https://github.com/qurator-spk/eynollah/actions/workflows/test-eynollah.yml) @@ -19,9 +20,11 @@ * Output in [PAGE-XML](https://github.com/PRImA-Research-Lab/PAGE-XML) * [OCR-D](https://github.com/qurator-spk/eynollah#use-as-ocr-d-processor) interface -:warning: Development is currently focused on achieving the best possible quality of results for a wide variety of historical documents and therefore processing can be very slow. We aim to improve this, but contributions are welcome. +:warning: Development is currently focused on achieving the best possible quality of results for a wide variety of +historical documents and therefore processing can be very slow. We aim to improve this, but contributions are welcome. ## Installation + Python `3.8-3.11` with Tensorflow `<2.13` on Linux are currently supported. For (limited) GPU support the CUDA toolkit needs to be installed. @@ -41,19 +44,40 @@ cd eynollah; pip install -e . Alternatively, you can run `make install` or `make install-dev` for editable installation. +To also install the dependencies for the OCR engines: + +``` +pip install "eynollah[OCR]" +# or +make install EXTRAS=OCR +``` + ## Models -Pre-trained models can be downloaded from [qurator-data.de](https://qurator-data.de/eynollah/) or [huggingface](https://huggingface.co/SBB?search_models=eynollah). +Pretrained models can be downloaded from [zenodo](https://zenodo.org/records/17194824) or [huggingface](https://huggingface.co/SBB?search_models=eynollah). For documentation on methods and models, have a look at [`models.md`](https://github.com/qurator-spk/eynollah/tree/main/docs/models.md). ## Train + In case you want to train your own model with Eynollah, have a look at [`train.md`](https://github.com/qurator-spk/eynollah/tree/main/docs/train.md). ## Usage -The command-line interface can be called like this: + +Eynollah supports five use cases: layout analysis (segmentation), binarization, +image enhancement, text recognition (OCR), and (trainable) reading order detection. + +### Layout Analysis + +The layout analysis module is responsible for detecting layouts, identifying text lines, and determining reading order +using both heuristic methods or a machine-based reading order detection model. + +Note that there are currently two supported ways for reading order detection: either as part of layout analysis based +on image input, or, currently under development, for given layout analysis results based on PAGE-XML data as input. + +The command-line interface for layout analysis can be called like this: ```sh -eynollah \ +eynollah layout \ -i | -di \ -o \ -m \ @@ -66,6 +90,7 @@ The following options can be used to further configure the processing: |-------------------|:-------------------------------------------------------------------------------| | `-fl` | full layout analysis including all steps and segmentation classes | | `-light` | lighter and faster but simpler method for main region detection and deskewing | +| `-tll` | this indicates the light textline and should be passed with light version | | `-tab` | apply table detection | | `-ae` | apply enhancement (the resulting image is saved to the output directory) | | `-as` | apply scaling | @@ -80,9 +105,51 @@ The following options can be used to further configure the processing: | `-sp ` | save cropped page image to this directory | | `-sa ` | save all (plot, enhanced/binary image, layout) to this directory | -If no option is set, the tool performs layout detection of main regions (background, text, images, separators and marginals). +If no option is set, the tool performs layout detection of main regions (background, text, images, separators +and marginals). The best output quality is produced when RGB images are used as input rather than greyscale or binarized images. +### Binarization + +The binarization module performs document image binarization using pretrained pixelwise segmentation models. + +The command-line interface for binarization of single image can be called like this: + +```sh +eynollah binarization \ + -i | -di \ + -o \ + -m \ +``` + +### OCR + +The OCR module performs text recognition from images using two main families of pretrained models: CNN-RNN–based OCR and Transformer-based OCR. + +The command-line interface for ocr can be called like this: + +```sh +eynollah ocr \ + -i | -di \ + -dx \ + -o \ + -m | --model_name \ +``` + +### Machine-based-reading-order + +The machine-based reading-order module employs a pretrained model to identify the reading order from layouts represented in PAGE-XML files. + +The command-line interface for machine based reading order can be called like this: + +```sh +eynollah machine-based-reading-order \ + -i | -di \ + -xml | -dx \ + -m \ + -o +``` + #### Use as OCR-D processor Eynollah ships with a CLI interface to be used as [OCR-D](https://ocr-d.de) [processor](https://ocr-d.de/en/spec/cli), @@ -90,8 +157,7 @@ formally described in [`ocrd-tool.json`](https://github.com/qurator-spk/eynollah In this case, the source image file group with (preferably) RGB images should be used as input like this: - ocrd-eynollah-segment -I OCR-D-IMG -O OCR-D-SEG -P models 2022-04-05 - + ocrd-eynollah-segment -I OCR-D-IMG -O OCR-D-SEG -P models eynollah_layout_v0_5_0 If the input file group is PAGE-XML (from a previous OCR-D workflow step), Eynollah behaves as follows: - existing regions are kept and ignored (i.e. in effect they might overlap segments from Eynollah results) @@ -103,15 +169,20 @@ If the input file group is PAGE-XML (from a previous OCR-D workflow step), Eynol (because some other preprocessing step was in effect like `denoised`), then the output PAGE-XML will be based on that as new top-level (`@imageFilename`) - - ocrd-eynollah-segment -I OCR-D-XYZ -O OCR-D-SEG -P models 2022-04-05 + ocrd-eynollah-segment -I OCR-D-XYZ -O OCR-D-SEG -P models eynollah_layout_v0_5_0 Still, in general, it makes more sense to add other workflow steps **after** Eynollah. +There is also an OCR-D processor for the binarization: + + ocrd-sbb-binarize -I OCR-D-IMG -O OCR-D-BIN -P models default-2021-03-09 + #### Additional documentation + Please check the [wiki](https://github.com/qurator-spk/eynollah/wiki). ## How to cite + If you find this tool useful in your work, please consider citing our paper: ```bibtex diff --git a/docs/models.md b/docs/models.md index ac563b0..3d296d5 100644 --- a/docs/models.md +++ b/docs/models.md @@ -1,5 +1,6 @@ # Models documentation -This suite of 14 models presents a document layout analysis (DLA) system for historical documents implemented by + +This suite of 15 models presents a document layout analysis (DLA) system for historical documents implemented by pixel-wise segmentation using a combination of a ResNet50 encoder with various U-Net decoders. In addition, heuristic methods are applied to detect marginals and to determine the reading order of text regions. @@ -23,6 +24,7 @@ See the flowchart below for the different stages and how they interact: ## Models ### Image enhancement + Model card: [Image Enhancement](https://huggingface.co/SBB/eynollah-enhancement) This model addresses image resolution, specifically targeting documents with suboptimal resolution. In instances where @@ -30,12 +32,14 @@ the detection of document layout exhibits inadequate performance, the proposed e the quality and clarity of the images, thus facilitating enhanced visual interpretation and analysis. ### Page extraction / border detection + Model card: [Page Extraction/Border Detection](https://huggingface.co/SBB/eynollah-page-extraction) A problem that can negatively affect OCR are black margins around a page caused by document scanning. A deep learning model helps to crop to the page borders by using a pixel-wise segmentation method. ### Column classification + Model card: [Column Classification](https://huggingface.co/SBB/eynollah-column-classifier) This model is a trained classifier that recognizes the number of columns in a document by use of a training set with @@ -43,6 +47,7 @@ manual classification of all documents into six classes with either one, two, th respectively. ### Binarization + Model card: [Binarization](https://huggingface.co/SBB/eynollah-binarization) This model is designed to tackle the intricate task of document image binarization, which involves segmentation of the @@ -52,6 +57,7 @@ capability of the model enables improved accuracy and reliability in subsequent enhanced document understanding and interpretation. ### Main region detection + Model card: [Main Region Detection](https://huggingface.co/SBB/eynollah-main-regions) This model has employed a different set of labels, including an artificial class specifically designed to encompass the @@ -61,6 +67,7 @@ during the inference phase. By incorporating this methodology, improved efficien model's ability to accurately identify and classify text regions within documents. ### Main region detection (with scaling augmentation) + Model card: [Main Region Detection (with scaling augmentation)](https://huggingface.co/SBB/eynollah-main-regions-aug-scaling) Utilizing scaling augmentation, this model leverages the capability to effectively segment elements of extremely high or @@ -69,12 +76,14 @@ categorizing and isolating such elements, thereby enhancing its overall performa documents with varying scale characteristics. ### Main region detection (with rotation augmentation) + Model card: [Main Region Detection (with rotation augmentation)](https://huggingface.co/SBB/eynollah-main-regions-aug-rotation) This model takes advantage of rotation augmentation. This helps the tool to segment the vertical text regions in a robust way. ### Main region detection (ensembled) + Model card: [Main Region Detection (ensembled)](https://huggingface.co/SBB/eynollah-main-regions-ensembled) The robustness of this model is attained through an ensembling technique that combines the weights from various epochs. @@ -82,16 +91,19 @@ By employing this approach, the model achieves a high level of resilience and st strengths of multiple epochs to enhance its overall performance and deliver consistent and reliable results. ### Full region detection (1,2-column documents) + Model card: [Full Region Detection (1,2-column documents)](https://huggingface.co/SBB/eynollah-full-regions-1column) This model deals with documents comprising of one and two columns. ### Full region detection (3,n-column documents) + Model card: [Full Region Detection (3,n-column documents)](https://huggingface.co/SBB/eynollah-full-regions-3pluscolumn) This model is responsible for detecting headers and drop capitals in documents with three or more columns. ### Textline detection + Model card: [Textline Detection](https://huggingface.co/SBB/eynollah-textline) The method for textline detection combines deep learning and heuristics. In the deep learning part, an image-to-image @@ -106,6 +118,7 @@ segmentation is first deskewed and then the textlines are separated with the sam textline bounding boxes. Later, the strap is rotated back into its original orientation. ### Textline detection (light) + Model card: [Textline Detection Light (simpler but faster method)](https://huggingface.co/SBB/eynollah-textline_light) The method for textline detection combines deep learning and heuristics. In the deep learning part, an image-to-image @@ -119,6 +132,7 @@ enhancing the model's ability to accurately identify and delineate individual te eliminates the need for additional heuristics in extracting textline contours. ### Table detection + Model card: [Table Detection](https://huggingface.co/SBB/eynollah-tables) The objective of this model is to perform table segmentation in historical document images. Due to the pixel-wise @@ -128,17 +142,21 @@ effectively identify and delineate tables within the historical document images, enabling subsequent analysis and interpretation. ### Image detection + Model card: [Image Detection](https://huggingface.co/SBB/eynollah-image-extraction) This model is used for the task of illustration detection only. ### Reading order detection + Model card: [Reading Order Detection]() TODO ## Heuristic methods + Additionally, some heuristic methods are employed to further improve the model predictions: + * After border detection, the largest contour is determined by a bounding box, and the image cropped to these coordinates. * For text region detection, the image is scaled up to make it easier for the model to detect background space between text regions. * A minimum area is defined for text regions in relation to the overall image dimensions, so that very small regions that are noise can be filtered out. diff --git a/docs/train.md b/docs/train.md index 9f44a63..47ad67b 100644 --- a/docs/train.md +++ b/docs/train.md @@ -1,4 +1,5 @@ # Training documentation + This aims to assist users in preparing training datasets, training models, and performing inference with trained models. We cover various use cases including pixel-wise segmentation, image classification, image enhancement, and machine-based reading order detection. For each use case, we provide guidance on how to generate the corresponding training dataset. @@ -11,6 +12,7 @@ The following three tasks can all be accomplished using the code in the * inference with the trained model ## Generate training dataset + The script `generate_gt_for_training.py` is used for generating training datasets. As the results of the following command demonstrates, the dataset generator provides three different commands: @@ -23,14 +25,19 @@ These three commands are: * pagexml2label ### image-enhancement + Generating a training dataset for image enhancement is quite straightforward. All that is needed is a set of high-resolution images. The training dataset can then be generated using the following command: -`python generate_gt_for_training.py image-enhancement -dis "dir of high resolution images" -dois "dir where degraded -images will be written" -dols "dir where the corresponding high resolution image will be written as label" -scs -"degrading scales json file"` +```sh +python generate_gt_for_training.py image-enhancement \ + -dis "dir of high resolution images" \ + -dois "dir where degraded images will be written" \ + -dols "dir where the corresponding high resolution image will be written as label" \ + -scs "degrading scales json file" +``` -The scales JSON file is a dictionary with a key named 'scales' and values representing scales smaller than 1. Images are +The scales JSON file is a dictionary with a key named `scales` and values representing scales smaller than 1. Images are downscaled based on these scales and then upscaled again to their original size. This process causes the images to lose resolution at different scales. The degraded images are used as input images, and the original high-resolution images serve as labels. The enhancement model can be trained with this generated dataset. The scales JSON file looks like this: @@ -42,6 +49,7 @@ serve as labels. The enhancement model can be trained with this generated datase ``` ### machine-based-reading-order + For machine-based reading order, we aim to determine the reading priority between two sets of text regions. The model's input is a three-channel image: the first and last channels contain information about each of the two text regions, while the middle channel encodes prominent layout elements necessary for reading order, such as separators and headers. @@ -52,10 +60,18 @@ For output images, it is necessary to specify the width and height. Additionally to filter out regions smaller than this minimum size. This minimum size is defined as the ratio of the text region area to the image area, with a default value of zero. To run the dataset generator, use the following command: -`python generate_gt_for_training.py machine-based-reading-order -dx "dir of GT xml files" -domi "dir where output images -will be written" -docl "dir where the labels will be written" -ih "height" -iw "width" -min "min area ratio"` +```shell +python generate_gt_for_training.py machine-based-reading-order \ + -dx "dir of GT xml files" \ + -domi "dir where output images will be written" \ + -docl "dir where the labels will be written" \ + -ih "height" \ + -iw "width" \ + -min "min area ratio" +``` ### pagexml2label + pagexml2label is designed to generate labels from GT page XML files for various pixel-wise segmentation use cases, including 'layout,' 'textline,' 'printspace,' 'glyph,' and 'word' segmentation. To train a pixel-wise segmentation model, we require images along with their corresponding labels. Our training script @@ -119,9 +135,13 @@ graphic region, "stamp" has its own class, while all other types are classified region" are also present in the label. However, other regions like "noise region" and "table region" will not be included in the label PNG file, even if they have information in the page XML files, as we chose not to include them. -`python generate_gt_for_training.py pagexml2label -dx "dir of GT xml files" -do "dir where output label png files will -be written" -cfg "custom config json file" -to "output type which has 2d and 3d. 2d is used for training and 3d is just -to visualise the labels" "` +```sh +python generate_gt_for_training.py pagexml2label \ + -dx "dir of GT xml files" \ + -do "dir where output label png files will be written" \ + -cfg "custom config json file" \ + -to "output type which has 2d and 3d. 2d is used for training and 3d is just to visualise the labels" +``` We have also defined an artificial class that can be added to the boundary of text region types or text lines. This key is called "artificial_class_on_boundary." If users want to apply this to certain text regions in the layout use case, @@ -169,12 +189,19 @@ in this scenario, since cropping will be applied to the label files, the directo provided to ensure that they are cropped in sync with the labels. This ensures that the correct images and labels required for training are obtained. The command should resemble the following: -`python generate_gt_for_training.py pagexml2label -dx "dir of GT xml files" -do "dir where output label png files will -be written" -cfg "custom config json file" -to "output type which has 2d and 3d. 2d is used for training and 3d is just -to visualise the labels" -ps -di "dir where the org images are located" -doi "dir where the cropped output images will -be written" ` +```sh +python generate_gt_for_training.py pagexml2label \ + -dx "dir of GT xml files" \ + -do "dir where output label png files will be written" \ + -cfg "custom config json file" \ + -to "output type which has 2d and 3d. 2d is used for training and 3d is just to visualise the labels" \ + -ps \ + -di "dir where the org images are located" \ + -doi "dir where the cropped output images will be written" +``` ## Train a model + ### classification For the classification use case, we haven't provided a ground truth generator, as it's unnecessary. For classification, @@ -225,7 +252,9 @@ And the "dir_eval" the same structure as train directory: The classification model can be trained using the following command line: -`python train.py with config_classification.json` +```sh +python train.py with config_classification.json +``` As evident in the example JSON file above, for classification, we utilize a "f1_threshold_classification" parameter. This parameter is employed to gather all models with an evaluation f1 score surpassing this threshold. Subsequently, @@ -276,6 +305,7 @@ The classification model can be trained like the classification case command lin ### Segmentation (Textline, Binarization, Page extraction and layout) and enhancement #### Parameter configuration for segmentation or enhancement usecases + The following parameter configuration can be applied to all segmentation use cases and enhancements. The augmentation, its sub-parameters, and continued training are defined only for segmentation use cases and enhancements, not for classification and machine-based reading order, as you can see in their example config files. @@ -355,6 +385,7 @@ command, similar to the process for classification and reading order: `python train.py with config_classification.json` #### Binarization + An example config json file for binarization can be like this: ```yaml @@ -550,6 +581,7 @@ For page segmentation (or printspace or border segmentation), the model needs to hence the patches parameter should be set to false. #### layout segmentation + An example config json file for layout segmentation with 5 classes (including background) can be like this: ```yaml @@ -605,26 +637,41 @@ An example config json file for layout segmentation with 5 classes (including ba ## Inference with the trained model ### classification + For conducting inference with a trained model, you simply need to execute the following command line, specifying the directory of the model and the image on which to perform inference: -`python inference.py -m "model dir" -i "image" ` +```sh +python inference.py -m "model dir" -i "image" +``` This will straightforwardly return the class of the image. ### machine based reading order + To infer the reading order using a reading order model, we need a page XML file containing layout information but without the reading order. We simply need to provide the model directory, the XML file, and the output directory. The new XML file with the added reading order will be written to the output directory with the same name. We need to run: -`python inference.py -m "model dir" -xml "page xml file" -o "output dir to write new xml with reading order" ` +```sh +python inference.py \ + -m "model dir" \ + -xml "page xml file" \ + -o "output dir to write new xml with reading order" +``` ### Segmentation (Textline, Binarization, Page extraction and layout) and enhancement For conducting inference with a trained model for segmentation and enhancement you need to run the following command line: -`python inference.py -m "model dir" -i "image" -p -s "output image" ` +```sh +python inference.py \ + -m "model dir" \ + -i "image" \ + -p \ + -s "output image" +``` Note that in the case of page extraction the -p flag is not needed. diff --git a/pyproject.toml b/pyproject.toml index 4da39ef..8a63543 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -46,7 +46,7 @@ optional-dependencies.test = {file = ["requirements-test.txt"]} where = ["src"] [tool.setuptools.package-data] -"*" = ["*.json", '*.yml', '*.xml', '*.xsd'] +"*" = ["*.json", '*.yml', '*.xml', '*.xsd', '*.ttf'] [tool.coverage.run] branch = true diff --git a/src/eynollah/Charis-Regular.ttf b/src/eynollah/Charis-Regular.ttf new file mode 100644 index 0000000..a4e75a4 Binary files /dev/null and b/src/eynollah/Charis-Regular.ttf differ diff --git a/src/eynollah/cli.py b/src/eynollah/cli.py index 9dc326d..93bb676 100644 --- a/src/eynollah/cli.py +++ b/src/eynollah/cli.py @@ -1,5 +1,6 @@ import sys import click +import logging from ocrd_utils import initLogging, getLevelName, getLogger from eynollah.eynollah import Eynollah, Eynollah_ocr from eynollah.sbb_binarize import SbbBinarizer @@ -12,22 +13,23 @@ def main(): @main.command() @click.option( - "--dir_xml", - "-dx", - help="directory of page-xml files", - type=click.Path(exists=True, file_okay=False), -) -@click.option( - "--xml_file", - "-xml", - help="xml filename", + "--input", + "-i", + help="PAGE-XML input filename", type=click.Path(exists=True, dir_okay=False), ) @click.option( - "--dir_out", - "-do", + "--dir_in", + "-di", + help="directory of PAGE-XML input files (instead of --input)", + type=click.Path(exists=True, file_okay=False), +) +@click.option( + "--out", + "-o", help="directory for output images", type=click.Path(exists=True, file_okay=False), + required=True, ) @click.option( "--model", @@ -36,53 +38,73 @@ def main(): type=click.Path(exists=True, file_okay=False), required=True, ) +@click.option( + "--log_level", + "-l", + type=click.Choice(['OFF', 'DEBUG', 'INFO', 'WARN', 'ERROR']), + help="Override log level globally to this", +) -def machine_based_reading_order(dir_xml, xml_file, dir_out, model): - raedingorder_object = machine_based_reading_order_on_layout(model, dir_out=dir_out, logger=getLogger('enhancement')) - - if dir_xml: - raedingorder_object.run(dir_in=dir_xml) - else: - raedingorder_object.run(xml_filename=xml_file) +def machine_based_reading_order(input, dir_in, out, model, log_level): + assert bool(input) != bool(dir_in), "Either -i (single input) or -di (directory) must be provided, but not both." + orderer = machine_based_reading_order_on_layout(model) + if log_level: + orderer.logger.setLevel(getLevelName(log_level)) + + orderer.run(xml_filename=input, + dir_in=dir_in, + dir_out=out, + ) @main.command() @click.option('--patches/--no-patches', default=True, help='by enabling this parameter you let the model to see the image in patches.') @click.option('--model_dir', '-m', type=click.Path(exists=True, file_okay=False), required=True, help='directory containing models for prediction') -@click.argument('input_image', required=False) -@click.argument('output_image', required=False) +@click.option( + "--input-image", "--image", + "-i", + help="input image filename", + type=click.Path(exists=True, dir_okay=False) +) @click.option( "--dir_in", "-di", - help="directory of input images", + help="directory of input images (instead of --image)", type=click.Path(exists=True, file_okay=False), ) @click.option( - "--dir_out", - "-do", - help="directory for output images", - type=click.Path(exists=True, file_okay=False), + "--output", + "-o", + help="output image (if using -i) or output image directory (if using -di)", + type=click.Path(file_okay=True, dir_okay=True), + required=True, ) -def binarization(patches, model_dir, input_image, output_image, dir_in, dir_out): - assert (dir_out is None) == (dir_in is None), "Options -di and -do are mutually dependent" - assert (input_image is None) == (output_image is None), "INPUT_IMAGE and OUTPUT_IMAGE are mutually dependent" - assert (dir_in is None) != (input_image is None), "Specify either -di and -do options, or INPUT_IMAGE and OUTPUT_IMAGE" - SbbBinarizer(model_dir).run(image_path=input_image, use_patches=patches, save=output_image, dir_in=dir_in, dir_out=dir_out) - +@click.option( + "--log_level", + "-l", + type=click.Choice(['OFF', 'DEBUG', 'INFO', 'WARN', 'ERROR']), + help="Override log level globally to this", +) +def binarization(patches, model_dir, input_image, dir_in, output, log_level): + assert bool(input_image) != bool(dir_in), "Either -i (single input) or -di (directory) must be provided, but not both." + binarizer = SbbBinarizer(model_dir) + if log_level: + binarizer.log.setLevel(getLevelName(log_level)) + binarizer.run(image_path=input_image, use_patches=patches, output=output, dir_in=dir_in) @main.command() @click.option( "--image", "-i", - help="image filename", + help="input image filename", type=click.Path(exists=True, dir_okay=False), ) @click.option( "--out", "-o", - help="directory to write output xml data", + help="directory for output PAGE-XML files", type=click.Path(exists=True, file_okay=False), required=True, ) @@ -95,7 +117,7 @@ def binarization(patches, model_dir, input_image, output_image, dir_in, dir_out) @click.option( "--dir_in", "-di", - help="directory of images", + help="directory of input images (instead of --image)", type=click.Path(exists=True, file_okay=False), ) @click.option( @@ -130,35 +152,34 @@ def binarization(patches, model_dir, input_image, output_image, dir_in, dir_out) ) def enhancement(image, out, overwrite, dir_in, model, num_col_upper, num_col_lower, save_org_scale, log_level): + assert bool(image) != bool(dir_in), "Either -i (single input) or -di (directory) must be provided, but not both." initLogging() - if log_level: - getLogger('enhancement').setLevel(getLevelName(log_level)) - assert image or dir_in, "Either a single image -i or a dir_in -di is required" - enhancer_object = Enhancer( + enhancer = Enhancer( model, - logger=getLogger('enhancement'), - dir_out=out, num_col_upper=num_col_upper, num_col_lower=num_col_lower, save_org_scale=save_org_scale, ) - if dir_in: - enhancer_object.run(dir_in=dir_in, overwrite=overwrite) - else: - enhancer_object.run(image_filename=image, overwrite=overwrite) + if log_level: + enhancer.logger.setLevel(getLevelName(log_level)) + enhancer.run(overwrite=overwrite, + dir_in=dir_in, + image_filename=image, + dir_out=out, + ) @main.command() @click.option( "--image", "-i", - help="image filename", + help="input image filename", type=click.Path(exists=True, dir_okay=False), ) @click.option( "--out", "-o", - help="directory to write output xml data", + help="directory for output PAGE-XML files", type=click.Path(exists=True, file_okay=False), required=True, ) @@ -171,7 +192,7 @@ def enhancement(image, out, overwrite, dir_in, model, num_col_upper, num_col_low @click.option( "--dir_in", "-di", - help="directory of images", + help="directory of input images (instead of --image)", type=click.Path(exists=True, file_okay=False), ) @click.option( @@ -338,17 +359,30 @@ def enhancement(image, out, overwrite, dir_in, model, num_col_upper, num_col_low is_flag=True, help="if this parameter set to true, this tool will ignore layout detection and reading order. It means that textline detection will be done within printspace and contours of textline will be written in xml output file.", ) +# TODO move to top-level CLI context @click.option( "--log_level", "-l", type=click.Choice(['OFF', 'DEBUG', 'INFO', 'WARN', 'ERROR']), - help="Override log level globally to this", + help="Override 'eynollah' log level globally to this", +) +# +@click.option( + "--setup-logging", + is_flag=True, + help="Setup a basic console logger", ) -def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_deskewed, save_all, extract_only_images, save_page, enable_plotting, allow_enhancement, curved_line, textline_light, full_layout, tables, right2left, input_binary, allow_scaling, headers_off, light_version, reading_order_machine_based, do_ocr, transformer_ocr, batch_size_ocr, num_col_upper, num_col_lower, threshold_art_class_textline, threshold_art_class_layout, skip_layout_and_reading_order, ignore_page_extraction, log_level): - initLogging() - if log_level: - getLogger('eynollah').setLevel(getLevelName(log_level)) +def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_deskewed, save_all, extract_only_images, save_page, enable_plotting, allow_enhancement, curved_line, textline_light, full_layout, tables, right2left, input_binary, allow_scaling, headers_off, light_version, reading_order_machine_based, do_ocr, transformer_ocr, batch_size_ocr, num_col_upper, num_col_lower, threshold_art_class_textline, threshold_art_class_layout, skip_layout_and_reading_order, ignore_page_extraction, log_level, setup_logging): + if setup_logging: + console_handler = logging.StreamHandler(sys.stdout) + console_handler.setLevel(logging.INFO) + formatter = logging.Formatter('%(message)s') + console_handler.setFormatter(formatter) + getLogger('eynollah').addHandler(console_handler) + getLogger('eynollah').setLevel(logging.INFO) + else: + initLogging() assert enable_plotting or not save_layout, "Plotting with -sl also requires -ep" assert enable_plotting or not save_deskewed, "Plotting with -sd also requires -ep" assert enable_plotting or not save_all, "Plotting with -sa also requires -ep" @@ -367,17 +401,10 @@ def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_ assert not extract_only_images or not tables, "Image extraction -eoi can not be set alongside tables -tab" assert not extract_only_images or not right2left, "Image extraction -eoi can not be set alongside right2left -r2l" assert not extract_only_images or not headers_off, "Image extraction -eoi can not be set alongside headers_off -ho" - assert image or dir_in, "Either a single image -i or a dir_in -di is required" + assert bool(image) != bool(dir_in), "Either -i (single input) or -di (directory) must be provided, but not both." eynollah = Eynollah( model, - logger=getLogger('eynollah'), - dir_out=out, - dir_of_cropped_images=save_images, extract_only_images=extract_only_images, - dir_of_layout=save_layout, - dir_of_deskewed=save_deskewed, - dir_of_all=save_all, - dir_save_page=save_page, enable_plotting=enable_plotting, allow_enhancement=allow_enhancement, curved_line=curved_line, @@ -400,56 +427,64 @@ def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_ threshold_art_class_textline=threshold_art_class_textline, threshold_art_class_layout=threshold_art_class_layout, ) - if dir_in: - eynollah.run(dir_in=dir_in, overwrite=overwrite) - else: - eynollah.run(image_filename=image, overwrite=overwrite) - + if log_level: + eynollah.logger.setLevel(getLevelName(log_level)) + eynollah.run(overwrite=overwrite, + image_filename=image, + dir_in=dir_in, + dir_out=out, + dir_of_cropped_images=save_images, + dir_of_layout=save_layout, + dir_of_deskewed=save_deskewed, + dir_of_all=save_all, + dir_save_page=save_page, + ) @main.command() @click.option( "--image", "-i", - help="image filename", + help="input image filename", type=click.Path(exists=True, dir_okay=False), ) +@click.option( + "--dir_in", + "-di", + help="directory of input images (instead of --image)", + type=click.Path(exists=True, file_okay=False), +) +@click.option( + "--dir_in_bin", + "-dib", + help="directory of binarized images (in addition to --dir_in for RGB images; filename stems must match the RGB image files, with '.png' suffix).\nPerform prediction using both RGB and binary images. (This does not necessarily improve results, however it may be beneficial for certain document images.)", + type=click.Path(exists=True, file_okay=False), +) +@click.option( + "--dir_xmls", + "-dx", + help="directory of input PAGE-XML files (in addition to --dir_in; filename stems must match the image files, with '.xml' suffix).", + type=click.Path(exists=True, file_okay=False), + required=True, +) +@click.option( + "--out", + "-o", + help="directory for output PAGE-XML files", + type=click.Path(exists=True, file_okay=False), + required=True, +) +@click.option( + "--dir_out_image_text", + "-doit", + help="directory for output images, newly rendered with predicted text", + type=click.Path(exists=True, file_okay=False), +) @click.option( "--overwrite", "-O", help="overwrite (instead of skipping) if output xml exists", is_flag=True, ) -@click.option( - "--dir_in", - "-di", - help="directory of images", - type=click.Path(exists=True, file_okay=False), -) -@click.option( - "--dir_in_bin", - "-dib", - help="directory of binarized images. This should be given if you want to do prediction based on both rgb and bin images. And all bin images are png files", - type=click.Path(exists=True, file_okay=False), -) -@click.option( - "--out", - "-o", - help="directory to write output xml data", - type=click.Path(exists=True, file_okay=False), - required=True, -) -@click.option( - "--dir_xmls", - "-dx", - help="directory of xmls", - type=click.Path(exists=True, file_okay=False), -) -@click.option( - "--dir_out_image_text", - "-doit", - help="directory of images with predicted text", - type=click.Path(exists=True, file_okay=False), -) @click.option( "--model", "-m", @@ -479,12 +514,6 @@ def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_ is_flag=True, help="if this parameter set to true, cropped textline images will not be masked with textline contour.", ) -@click.option( - "--prediction_with_both_of_rgb_and_bin", - "-brb/-nbrb", - is_flag=True, - help="If this parameter is set to True, the prediction will be performed using both RGB and binary images. However, this does not necessarily improve results; it may be beneficial for certain document images.", -) @click.option( "--batch_size", "-bs", @@ -507,37 +536,36 @@ def layout(image, out, overwrite, dir_in, model, save_images, save_layout, save_ help="Override log level globally to this", ) -def ocr(image, overwrite, dir_in, dir_in_bin, out, dir_xmls, dir_out_image_text, model, model_name, tr_ocr, export_textline_images_and_text, do_not_mask_with_textline_contour, prediction_with_both_of_rgb_and_bin, batch_size, dataset_abbrevation, min_conf_value_of_textline_text, log_level): +def ocr(image, dir_in, dir_in_bin, dir_xmls, out, dir_out_image_text, overwrite, model, model_name, tr_ocr, export_textline_images_and_text, do_not_mask_with_textline_contour, batch_size, dataset_abbrevation, min_conf_value_of_textline_text, log_level): initLogging() - if log_level: - getLogger('eynollah').setLevel(getLevelName(log_level)) - assert not model or not model_name, "model directory -m can not be set alongside specific model name --model_name" + assert bool(model) != bool(model_name), "Either -m (model directory) or --model_name (specific model name) must be provided." assert not export_textline_images_and_text or not tr_ocr, "Exporting textline and text -etit can not be set alongside transformer ocr -tr_ocr" assert not export_textline_images_and_text or not model, "Exporting textline and text -etit can not be set alongside model -m" assert not export_textline_images_and_text or not batch_size, "Exporting textline and text -etit can not be set alongside batch size -bs" assert not export_textline_images_and_text or not dir_in_bin, "Exporting textline and text -etit can not be set alongside directory of bin images -dib" assert not export_textline_images_and_text or not dir_out_image_text, "Exporting textline and text -etit can not be set alongside directory of images with predicted text -doit" - assert not export_textline_images_and_text or not prediction_with_both_of_rgb_and_bin, "Exporting textline and text -etit can not be set alongside prediction with both rgb and bin -brb" - assert (bool(image) ^ bool(dir_in)), "Either -i (single image) or -di (directory) must be provided, but not both." + assert bool(image) != bool(dir_in), "Either -i (single image) or -di (directory) must be provided, but not both." eynollah_ocr = Eynollah_ocr( - image_filename=image, - dir_xmls=dir_xmls, - dir_out_image_text=dir_out_image_text, - dir_in=dir_in, - dir_in_bin=dir_in_bin, - dir_out=out, dir_models=model, model_name=model_name, tr_ocr=tr_ocr, export_textline_images_and_text=export_textline_images_and_text, do_not_mask_with_textline_contour=do_not_mask_with_textline_contour, - prediction_with_both_of_rgb_and_bin=prediction_with_both_of_rgb_and_bin, batch_size=batch_size, pref_of_dataset=dataset_abbrevation, min_conf_value_of_textline_text=min_conf_value_of_textline_text, ) - eynollah_ocr.run(overwrite=overwrite) + if log_level: + eynollah_ocr.logger.setLevel(getLevelName(log_level)) + eynollah_ocr.run(overwrite=overwrite, + dir_in=dir_in, + dir_in_bin=dir_in_bin, + image_filename=image, + dir_xmls=dir_xmls, + dir_out_image_text=dir_out_image_text, + dir_out=out, + ) if __name__ == "__main__": main() diff --git a/src/eynollah/eynollah.py b/src/eynollah/eynollah.py index 574d823..5e8412e 100644 --- a/src/eynollah/eynollah.py +++ b/src/eynollah/eynollah.py @@ -1,4 +1,4 @@ -# pylint: disable=no-member,invalid-name,line-too-long,missing-function-docstring,missing-class-docstring,too-many-branches +#run_single# pylint: disable=no-member,invalid-name,line-too-long,missing-function-docstring,missing-class-docstring,too-many-branches # pylint: disable=too-many-locals,wrong-import-position,too-many-lines,too-many-statements,chained-comparison,fixme,broad-except,c-extension-no-member # pylint: disable=too-many-public-methods,too-many-arguments,too-many-instance-attributes,too-many-public-methods, # pylint: disable=consider-using-enumerate @@ -6,7 +6,13 @@ document layout analysis (segmentation) with output in PAGE-XML """ -from logging import Logger +# cannot use importlib.resources until we move to 3.9+ forimportlib.resources.files +import sys +if sys.version_info < (3, 10): + import importlib_resources +else: + import importlib.resources as importlib_resources + from difflib import SequenceMatcher as sq from PIL import Image, ImageDraw, ImageFont import math @@ -108,6 +114,7 @@ from .utils.drop_capitals import ( from .utils.marginals import get_marginals from .utils.resize import resize_image from .utils import ( + is_image_filename, boosting_headers_by_longshot_region_segmentation, crop_image_inside_box, find_num_col, @@ -191,13 +198,7 @@ class Eynollah: def __init__( self, dir_models : str, - dir_out : Optional[str] = None, - dir_of_cropped_images : Optional[str] = None, extract_only_images : bool =False, - dir_of_layout : Optional[str] = None, - dir_of_deskewed : Optional[str] = None, - dir_of_all : Optional[str] = None, - dir_save_page : Optional[str] = None, enable_plotting : bool = False, allow_enhancement : bool = False, curved_line : bool = False, @@ -219,19 +220,14 @@ class Eynollah: threshold_art_class_layout: Optional[float] = None, threshold_art_class_textline: Optional[float] = None, skip_layout_and_reading_order : bool = False, - logger : Optional[Logger] = None, ): + self.logger = getLogger('eynollah') + self.plotter = None + if skip_layout_and_reading_order: textline_light = True self.light_version = light_version - self.dir_out = dir_out - self.dir_of_all = dir_of_all - self.dir_save_page = dir_save_page self.reading_order_machine_based = reading_order_machine_based - self.dir_of_deskewed = dir_of_deskewed - self.dir_of_deskewed = dir_of_deskewed - self.dir_of_cropped_images=dir_of_cropped_images - self.dir_of_layout=dir_of_layout self.enable_plotting = enable_plotting self.allow_enhancement = allow_enhancement self.curved_line = curved_line @@ -267,10 +263,6 @@ class Eynollah: else: self.threshold_art_class_textline = 0.1 - self.logger = logger if logger else getLogger('eynollah') - # for parallelization of CPU-intensive tasks: - self.executor = ProcessPoolExecutor(max_workers=cpu_count(), timeout=1200) - atexit.register(self.executor.shutdown) self.dir_models = dir_models self.model_dir_of_enhancement = dir_models + "/eynollah-enhancement_20210425" self.model_dir_of_binarization = dir_models + "/eynollah-binarization_20210425" @@ -325,7 +317,14 @@ class Eynollah: self.model_table_dir = dir_models + "/modelens_table_0t4_201124" else: self.model_table_dir = dir_models + "/eynollah-tables_20210319" + + + t_start = time.time() + # for parallelization of CPU-intensive tasks: + self.executor = ProcessPoolExecutor(max_workers=cpu_count(), timeout=1200) + atexit.register(self.executor.shutdown) + # #gpu_options = tf.compat.v1.GPUOptions(allow_growth=True) # #gpu_options = tf.compat.v1.GPUOptions(per_process_gpu_memory_fraction=7.7, allow_growth=True) # #session = tf.compat.v1.Session(config=tf.compat.v1.ConfigProto(gpu_options=gpu_options)) @@ -339,7 +338,9 @@ class Eynollah: tf.config.experimental.set_memory_growth(device, True) except: self.logger.warning("no GPU device available") - + + self.logger.info("Loading models...") + self.model_page = self.our_load_model(self.model_page_dir) self.model_classifier = self.our_load_model(self.model_dir_of_col_classifier) self.model_bin = self.our_load_model(self.model_dir_of_binarization) @@ -392,6 +393,8 @@ class Eynollah: if self.tables: self.model_table = self.our_load_model(self.model_table_dir) + + self.logger.info(f"Model initialization complete ({time.time() - t_start:.1f}s)") def cache_images(self, image_filename=None, image_pil=None, dpi=None): ret = {} @@ -415,21 +418,11 @@ class Eynollah: if dpi is not None: self.dpi = dpi - def reset_file_name_dir(self, image_filename): + def reset_file_name_dir(self, image_filename, dir_out): t_c = time.time() self.cache_images(image_filename=image_filename) - - self.plotter = None if not self.enable_plotting else EynollahPlotter( - dir_out=self.dir_out, - dir_of_all=self.dir_of_all, - dir_save_page=self.dir_save_page, - dir_of_deskewed=self.dir_of_deskewed, - dir_of_cropped_images=self.dir_of_cropped_images, - dir_of_layout=self.dir_of_layout, - image_filename_stem=Path(Path(image_filename).name).stem) - self.writer = EynollahXmlWriter( - dir_out=self.dir_out, + dir_out=dir_out, image_filename=image_filename, curved_line=self.curved_line, textline_light = self.textline_light) @@ -1746,12 +1739,85 @@ class Eynollah: prediction_regions = resize_image(prediction_regions, img_height_h, img_width_h) self.logger.debug("exit extract_text_regions") return prediction_regions, prediction_regions2 + + def get_textlines_of_a_textregion_sorted(self, textlines_textregion, cx_textline, cy_textline, w_h_textline): + N = len(cy_textline) + if N==0: + return [] + + diff_cy = np.abs( np.diff(sorted(cy_textline)) ) + diff_cx = np.abs(np.diff(sorted(cx_textline)) ) + + if len(diff_cy)>0: + mean_y_diff = np.mean(diff_cy) + mean_x_diff = np.mean(diff_cx) + count_hor = np.count_nonzero(np.array(w_h_textline) > 1) + count_ver = len(w_h_textline) - count_hor + + else: + mean_y_diff = 0 + mean_x_diff = 0 + count_hor = 1 + count_ver = 0 + + + if count_hor >= count_ver: + row_threshold = mean_y_diff / 1.5 if mean_y_diff > 0 else 10 + + indices_sorted_by_y = sorted(range(N), key=lambda i: cy_textline[i]) + + rows = [] + current_row = [indices_sorted_by_y[0]] + for i in range(1, N): + current_idx = indices_sorted_by_y[i] + prev_idx = current_row[0] + if abs(cy_textline[current_idx] - cy_textline[prev_idx]) <= row_threshold: + current_row.append(current_idx) + else: + rows.append(current_row) + current_row = [current_idx] + rows.append(current_row) + + sorted_textlines = [] + for row in rows: + row_sorted = sorted(row, key=lambda i: cx_textline[i]) + for idx in row_sorted: + sorted_textlines.append(textlines_textregion[idx]) + + else: + row_threshold = mean_x_diff / 1.5 if mean_x_diff > 0 else 10 + indices_sorted_by_x = sorted(range(N), key=lambda i: cx_textline[i]) + + rows = [] + current_row = [indices_sorted_by_x[0]] + + for i in range(1, N): + current_idy = indices_sorted_by_x[i] + prev_idy = current_row[0] + if abs(cx_textline[current_idy] - cx_textline[prev_idy] ) <= row_threshold: + current_row.append(current_idy) + else: + rows.append(current_row) + current_row = [current_idy] + rows.append(current_row) + + sorted_textlines = [] + for row in rows: + row_sorted = sorted(row , key=lambda i: cy_textline[i]) + for idy in row_sorted: + sorted_textlines.append(textlines_textregion[idy]) + + return sorted_textlines + + def get_slopes_and_deskew_new_light2(self, contours, contours_par, textline_mask_tot, image_page_rotated, boxes, slope_deskew): polygons_of_textlines = return_contours_of_interested_region(textline_mask_tot,1,0.00001) M_main_tot = [cv2.moments(polygons_of_textlines[j]) for j in range(len(polygons_of_textlines))] + + w_h_textlines = [cv2.boundingRect(polygons_of_textlines[i])[2:] for i in range(len(polygons_of_textlines))] cx_main_tot = [(M_main_tot[j]["m10"] / (M_main_tot[j]["m00"] + 1e-32)) for j in range(len(M_main_tot))] cy_main_tot = [(M_main_tot[j]["m01"] / (M_main_tot[j]["m00"] + 1e-32)) for j in range(len(M_main_tot))] @@ -1766,8 +1832,13 @@ class Eynollah: results = np.array(results) indexes_in = args_textlines[results==1] textlines_ins = [polygons_of_textlines[ind] for ind in indexes_in] + cx_textline_in = [cx_main_tot[ind] for ind in indexes_in] + cy_textline_in = [cy_main_tot[ind] for ind in indexes_in] + w_h_textlines_in = [w_h_textlines[ind][0] / float(w_h_textlines[ind][1]) for ind in indexes_in] - all_found_textline_polygons.append(textlines_ins[::-1]) + textlines_ins = self.get_textlines_of_a_textregion_sorted(textlines_ins, cx_textline_in, cy_textline_in, w_h_textlines_in) + + all_found_textline_polygons.append(textlines_ins)#[::-1]) slopes.append(slope_deskew) _, crop_coor = crop_image_inside_box(boxes[index],image_page_rotated) @@ -2174,6 +2245,7 @@ class Eynollah: ##mask_texts_only = cv2.dilate(mask_texts_only, KERNEL, iterations=1) mask_texts_only = cv2.dilate(mask_texts_only, kernel=np.ones((2,2), np.uint8), iterations=1) + mask_images_only=(prediction_regions_org[:,:] ==2)*1 polygons_lines_xml, hir_lines_xml = return_contours_of_image(mask_lines_only) @@ -2209,20 +2281,18 @@ class Eynollah: text_regions_p_true[:,:][mask_images_only[:,:] == 1] = 2 text_regions_p_true = cv2.fillPoly(text_regions_p_true, pts = polygons_of_only_texts, color=(1,1,1)) - #plt.imshow(textline_mask_tot_ea) #plt.show() textline_mask_tot_ea[(text_regions_p_true==0) | (text_regions_p_true==4) ] = 0 - #plt.imshow(textline_mask_tot_ea) #plt.show() #print("inside 4 ", time.time()-t_in) self.logger.debug("exit get_regions_light_v") - return text_regions_p_true, erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin, confidence_matrix + return text_regions_p_true, erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin, confidence_matrix, polygons_of_only_texts else: img_bin = resize_image(img_bin,img_height_h, img_width_h ) self.logger.debug("exit get_regions_light_v") - return None, erosion_hurts, None, textline_mask_tot_ea, img_bin, None + return None, erosion_hurts, None, textline_mask_tot_ea, img_bin, None, None def get_regions_from_xy_2models(self,img,is_image_enhanced, num_col_classifier): self.logger.debug("enter get_regions_from_xy_2models") @@ -2315,7 +2385,7 @@ class Eynollah: text_regions_p_true=cv2.fillPoly(text_regions_p_true,pts=polygons_of_only_texts, color=(1,1,1)) self.logger.debug("exit get_regions_from_xy_2models") - return text_regions_p_true, erosion_hurts, polygons_lines_xml + return text_regions_p_true, erosion_hurts, polygons_lines_xml, polygons_of_only_texts except: if self.input_binary: prediction_bin = np.copy(img_org) @@ -2365,7 +2435,7 @@ class Eynollah: erosion_hurts = True self.logger.debug("exit get_regions_from_xy_2models") - return text_regions_p_true, erosion_hurts, polygons_lines_xml + return text_regions_p_true, erosion_hurts, polygons_lines_xml, polygons_of_only_texts def do_order_of_regions_full_layout( self, contours_only_text_parent, contours_only_text_parent_h, boxes, textline_mask_tot): @@ -4517,27 +4587,68 @@ class Eynollah: return ordered_left_marginals, ordered_right_marginals, ordered_left_marginals_textline, ordered_right_marginals_textline, ordered_left_marginals_bbox, ordered_right_marginals_bbox, ordered_left_slopes_marginals, ordered_right_slopes_marginals - def run(self, image_filename : Optional[str] = None, dir_in : Optional[str] = None, overwrite : bool = False): + def run(self, + overwrite: bool = False, + image_filename: Optional[str] = None, + dir_in: Optional[str] = None, + dir_out: Optional[str] = None, + dir_of_cropped_images: Optional[str] = None, + dir_of_layout: Optional[str] = None, + dir_of_deskewed: Optional[str] = None, + dir_of_all: Optional[str] = None, + dir_save_page: Optional[str] = None, + ): """ Get image and scales, then extract the page of scanned image """ self.logger.debug("enter run") t0_tot = time.time() + # Log enabled features directly + enabled_modes = [] + if self.light_version: + enabled_modes.append("Light version") + if self.textline_light: + enabled_modes.append("Light textline detection") + if self.full_layout: + enabled_modes.append("Full layout analysis") + if self.ocr: + enabled_modes.append("OCR") + if self.tables: + enabled_modes.append("Table detection") + if enabled_modes: + self.logger.info("Enabled modes: " + ", ".join(enabled_modes)) + if self.enable_plotting: + self.logger.info("Saving debug plots") + if dir_of_cropped_images: + self.logger.info(f"Saving cropped images to: {dir_of_cropped_images}") + if dir_of_layout: + self.logger.info(f"Saving layout plots to: {dir_of_layout}") + if dir_of_deskewed: + self.logger.info(f"Saving deskewed images to: {dir_of_deskewed}") + if dir_in: - self.ls_imgs = os.listdir(dir_in) - self.ls_imgs = [ind_img for ind_img in self.ls_imgs if ind_img.endswith('.jpg') or ind_img.endswith('.jpeg') or ind_img.endswith('.png') or ind_img.endswith('.tif') or ind_img.endswith('.tiff') or ind_img.endswith('.JPG') or ind_img.endswith('.JPEG') or ind_img.endswith('.TIF') or ind_img.endswith('.TIFF') or ind_img.endswith('.PNG')] + ls_imgs = [os.path.join(dir_in, image_filename) + for image_filename in filter(is_image_filename, + os.listdir(dir_in))] elif image_filename: - self.ls_imgs = [image_filename] + ls_imgs = [image_filename] else: raise ValueError("run requires either a single image filename or a directory") - for img_filename in self.ls_imgs: - print(img_filename, 'img_filename') + for img_filename in ls_imgs: self.logger.info(img_filename) t0 = time.time() - self.reset_file_name_dir(os.path.join(dir_in or "", img_filename)) + self.reset_file_name_dir(img_filename, dir_out) + if self.enable_plotting: + self.plotter = EynollahPlotter(dir_out=dir_out, + dir_of_all=dir_of_all, + dir_save_page=dir_save_page, + dir_of_deskewed=dir_of_deskewed, + dir_of_cropped_images=dir_of_cropped_images, + dir_of_layout=dir_of_layout, + image_filename_stem=Path(image_filename).stem) #print("text region early -11 in %.1fs", time.time() - t0) if os.path.exists(self.writer.output_filename): if overwrite: @@ -4548,19 +4659,30 @@ class Eynollah: pcgts = self.run_single() self.logger.info("Job done in %.1fs", time.time() - t0) - #print("Job done in %.1fs" % (time.time() - t0)) self.writer.write_pagexml(pcgts) if dir_in: self.logger.info("All jobs done in %.1fs", time.time() - t0_tot) - print("all Job done in %.1fs", time.time() - t0_tot) def run_single(self): t0 = time.time() - img_res, is_image_enhanced, num_col_classifier, num_column_is_classified = self.run_enhancement(self.light_version) - self.logger.info("Enhancing took %.1fs ", time.time() - t0) + + self.logger.info(f"Processing file: {self.writer.image_filename}") + self.logger.info("Step 1/5: Image Enhancement") + img_res, is_image_enhanced, num_col_classifier, num_column_is_classified = self.run_enhancement(self.light_version) + + self.logger.info(f"Image: {self.image.shape[1]}x{self.image.shape[0]}, {self.dpi} DPI, {num_col_classifier} columns") + if is_image_enhanced: + self.logger.info("Enhancement applied") + + self.logger.info(f"Enhancement complete ({time.time() - t0:.1f}s)") + + + # Image Extraction Mode if self.extract_only_images: + self.logger.info("Step 2/5: Image Extraction Mode") + text_regions_p_1, erosion_hurts, polygons_lines_xml, polygons_of_images, image_page, page_coord, cont_page = \ self.get_regions_light_v_extract_only_images(img_res, is_image_enhanced, num_col_classifier) pcgts = self.writer.build_pagexml_no_full_layout( @@ -4569,10 +4691,16 @@ class Eynollah: cont_page, [], []) if self.plotter: self.plotter.write_images_into_directory(polygons_of_images, image_page) + + self.logger.info("Image extraction complete") return pcgts + # Basic Processing Mode if self.skip_layout_and_reading_order: - _ ,_, _, textline_mask_tot_ea, img_bin_light, _ = \ + self.logger.info("Step 2/5: Basic Processing Mode") + self.logger.info("Skipping layout analysis and reading order detection") + + _ ,_, _, textline_mask_tot_ea, img_bin_light, _,_= \ self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier, skip_layout_and_reading_order=self.skip_layout_and_reading_order) @@ -4585,7 +4713,14 @@ class Eynollah: all_found_textline_polygons = filter_contours_area_of_image( textline_mask_tot_ea, cnt_clean_rot_raw, hir_on_cnt_clean_rot, max_area=1, min_area=0.00001) - all_found_textline_polygons = all_found_textline_polygons[::-1] + M_main_tot = [cv2.moments(all_found_textline_polygons[j]) + for j in range(len(all_found_textline_polygons))] + w_h_textlines = [cv2.boundingRect(all_found_textline_polygons[j])[2:] for j in range(len(all_found_textline_polygons))] + w_h_textlines = [w_h_textlines[j][0] / float(w_h_textlines[j][1]) for j in range(len(w_h_textlines))] + cx_main_tot = [(M_main_tot[j]["m10"] / (M_main_tot[j]["m00"] + 1e-32)) for j in range(len(M_main_tot))] + cy_main_tot = [(M_main_tot[j]["m01"] / (M_main_tot[j]["m00"] + 1e-32)) for j in range(len(M_main_tot))] + + all_found_textline_polygons = self.get_textlines_of_a_textregion_sorted(all_found_textline_polygons, cx_main_tot, cy_main_tot, w_h_textlines)#all_found_textline_polygons[::-1] all_found_textline_polygons=[ all_found_textline_polygons ] @@ -4623,15 +4758,19 @@ class Eynollah: all_found_textline_polygons, page_coord, polygons_of_images, polygons_of_marginals_left, polygons_of_marginals_right, all_found_textline_polygons_marginals_left, all_found_textline_polygons_marginals_right, all_box_coord_marginals_left, all_box_coord_marginals_right, slopes, slopes_marginals_left, slopes_marginals_right, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines=ocr_all_textlines, conf_contours_textregion=conf_contours_textregions, skip_layout_reading_order=self.skip_layout_and_reading_order) + self.logger.info("Basic processing complete") return pcgts #print("text region early -1 in %.1fs", time.time() - t0) t1 = time.time() + self.logger.info("Step 2/5: Layout Analysis") + if self.light_version: - text_regions_p_1 ,erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin_light, confidence_matrix = \ + self.logger.info("Using light version processing") + text_regions_p_1 ,erosion_hurts, polygons_lines_xml, textline_mask_tot_ea, img_bin_light, confidence_matrix, polygons_text_early = \ self.get_regions_light_v(img_res, is_image_enhanced, num_col_classifier) #print("text region early -2 in %.1fs", time.time() - t0) - + if num_col_classifier == 1 or num_col_classifier ==2: if num_col_classifier == 1: img_w_new = 1000 @@ -4653,24 +4792,27 @@ class Eynollah: #self.logger.info("run graphics %.1fs ", time.time() - t1t) #print("text region early -3 in %.1fs", time.time() - t0) textline_mask_tot_ea_org = np.copy(textline_mask_tot_ea) - #print("text region early -4 in %.1fs", time.time() - t0) + else: - text_regions_p_1 ,erosion_hurts, polygons_lines_xml = \ + text_regions_p_1 ,erosion_hurts, polygons_lines_xml, polygons_text_early = \ self.get_regions_from_xy_2models(img_res, is_image_enhanced, num_col_classifier) - self.logger.info("Textregion detection took %.1fs ", time.time() - t1) + self.logger.info(f"Textregion detection took {time.time() - t1:.1f}s") confidence_matrix = np.zeros((text_regions_p_1.shape[:2])) t1 = time.time() num_col, num_col_classifier, img_only_regions, page_coord, image_page, mask_images, mask_lines, \ text_regions_p_1, cont_page, table_prediction = \ self.run_graphics_and_columns(text_regions_p_1, num_col_classifier, num_column_is_classified, erosion_hurts) - self.logger.info("Graphics detection took %.1fs ", time.time() - t1) + self.logger.info(f"Graphics detection took {time.time() - t1:.1f}s") #self.logger.info('cont_page %s', cont_page) #plt.imshow(table_prediction) #plt.show() - if not num_col: - self.logger.info("No columns detected, outputting an empty PAGE-XML") + self.logger.info(f"Layout analysis complete ({time.time() - t1:.1f}s)") + + if not num_col and len(polygons_text_early) == 0: + self.logger.info("No columns detected - generating empty PAGE-XML") + pcgts = self.writer.build_pagexml_no_full_layout( [], page_coord, [], [], [], [], [], [], [], [], [], [], [], [], [], [], cont_page, [], []) @@ -4680,10 +4822,12 @@ class Eynollah: t1 = time.time() if not self.light_version: textline_mask_tot_ea = self.run_textline(image_page) - self.logger.info("textline detection took %.1fs", time.time() - t1) + self.logger.info(f"Textline detection took {time.time() - t1:.1f}s") t1 = time.time() slope_deskew, slope_first = self.run_deskew(textline_mask_tot_ea) - self.logger.info("deskewing took %.1fs", time.time() - t1) + if np.abs(slope_deskew) > 0.01: # Only log if there is significant skew + self.logger.info(f"Applied deskew correction: {slope_deskew:.2f} degrees") + self.logger.info(f"Deskewing took {time.time() - t1:.1f}s") elif num_col_classifier in (1,2): org_h_l_m = textline_mask_tot_ea.shape[0] org_w_l_m = textline_mask_tot_ea.shape[1] @@ -4703,6 +4847,22 @@ class Eynollah: textline_mask_tot, text_regions_p, image_page_rotated = \ self.run_marginals(image_page, textline_mask_tot_ea, mask_images, mask_lines, num_col_classifier, slope_deskew, text_regions_p_1, table_prediction) + + + if image_page.shape[0]!=0 and image_page.shape[1]!=0: + # if ratio of text regions to page area is smaller that 0.3, deskew angle is not aloowed to exceed 45 + if ( ( text_regions_p[:,:]==1).sum() + (text_regions_p[:,:]==4).sum() ) / float(image_page.shape[0]*image_page.shape[1] ) <= 0.3 and abs(slope_deskew) > 45: + slope_deskew = 0 + + if (text_regions_p[:,:]==1).sum() == 0: + text_regions_p[:,:][text_regions_p[:,:]==4] = 1 + + self.logger.info("Step 3/5: Text Line Detection") + + if self.curved_line: + self.logger.info("Mode: Curved line detection") + elif self.textline_light: + self.logger.info("Mode: Light detection") if self.light_version and num_col_classifier in (1,2): image_page = resize_image(image_page,org_h_l_m, org_w_l_m ) @@ -4713,8 +4873,7 @@ class Eynollah: table_prediction = resize_image(table_prediction,org_h_l_m, org_w_l_m ) image_page_rotated = resize_image(image_page_rotated,org_h_l_m, org_w_l_m ) - self.logger.info("detection of marginals took %.1fs", time.time() - t1) - #print("text region early 2 marginal in %.1fs", time.time() - t0) + self.logger.info(f"Detection of marginals took {time.time() - t1:.1f}s") ## birdan sora chock chakir t1 = time.time() if not self.full_layout: @@ -4743,6 +4902,8 @@ class Eynollah: ###min_con_area = 0.000005 contours_only_text, hir_on_text = return_contours_of_image(text_only) contours_only_text_parent = return_parent_contours(contours_only_text, hir_on_text) + + if len(contours_only_text_parent) > 0: areas_cnt_text = np.array([cv2.contourArea(c) for c in contours_only_text_parent]) areas_cnt_text = areas_cnt_text / float(text_only.shape[0] * text_only.shape[1]) @@ -4812,7 +4973,7 @@ class Eynollah: cx_bigest_d_big[0] = cx_bigest_d[ind_largest] cy_biggest_d_big[0] = cy_biggest_d[ind_largest] except Exception as why: - self.logger.error(why) + self.logger.error(str(why)) (h, w) = text_only.shape[:2] center = (w // 2.0, h // 2.0) @@ -4844,7 +5005,9 @@ class Eynollah: contours_only_text_parent_d_ordered = [] contours_only_text_parent_d = [] #contours_only_text_parent = [] - + + boxes_marginals, _ = get_text_region_boxes_by_given_contours(polygons_of_marginals) + if not len(contours_only_text_parent): # stop early empty_marginals = [[]] * len(polygons_of_marginals) @@ -4880,7 +5043,6 @@ class Eynollah: contours_only_text_parent, self.image, slope_first, confidence_matrix, map=self.executor.map) #print("text region early 4 in %.1fs", time.time() - t0) boxes_text, _ = get_text_region_boxes_by_given_contours(contours_only_text_parent) - boxes_marginals, _ = get_text_region_boxes_by_given_contours(polygons_of_marginals) #print("text region early 5 in %.1fs", time.time() - t0) ## birdan sora chock chakir if not self.curved_line: @@ -5034,6 +5196,15 @@ class Eynollah: t_order = time.time() if self.full_layout: + self.logger.info("Step 4/5: Reading Order Detection") + + if self.reading_order_machine_based: + self.logger.info("Using machine-based detection") + if self.right2left: + self.logger.info("Right-to-left mode enabled") + if self.headers_off: + self.logger.info("Headers ignored in reading order") + if self.reading_order_machine_based: tror = time.time() order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model( @@ -5045,9 +5216,16 @@ class Eynollah: else: order_text_new, id_of_texts_tot = self.do_order_of_regions( contours_only_text_parent_d_ordered, contours_only_text_parent_h_d_ordered, boxes_d, textline_mask_tot_d) - self.logger.info("detection of reading order took %.1fs", time.time() - t_order) + self.logger.info(f"Detection of reading order took {time.time() - t_order:.1f}s") if self.ocr and not self.tr: + self.logger.info("Step 4.5/5: OCR Processing") + + if torch.cuda.is_available(): + self.logger.info("Using GPU acceleration") + else: + self.logger.info("Using CPU processing") + gc.collect() if len(all_found_textline_polygons)>0: ocr_all_textlines = return_rnn_cnn_ocr_of_given_textlines(image_page, all_found_textline_polygons, self.prediction_model, self.b_s_ocr, self.num_to_char, self.textline_light, self.curved_line) @@ -5079,15 +5257,28 @@ class Eynollah: ocr_all_textlines_marginals_right = None ocr_all_textlines_h = None ocr_all_textlines_drop = None + + self.logger.info("Step 5/5: Output Generation") + pcgts = self.writer.build_pagexml_full_layout( contours_only_text_parent, contours_only_text_parent_h, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_found_textline_polygons_h, all_box_coord, all_box_coord_h, polygons_of_images, contours_tables, polygons_of_drop_capitals, polygons_of_marginals_left, polygons_of_marginals_right, all_found_textline_polygons_marginals_left, all_found_textline_polygons_marginals_right, all_box_coord_marginals_left, all_box_coord_marginals_right, slopes, slopes_h, slopes_marginals_left, slopes_marginals_right, cont_page, polygons_lines_xml, ocr_all_textlines, ocr_all_textlines_h, ocr_all_textlines_marginals_left, ocr_all_textlines_marginals_right, ocr_all_textlines_drop, conf_contours_textregions, conf_contours_textregions_h) + return pcgts contours_only_text_parent_h = None + self.logger.info("Step 4/5: Reading Order Detection") + + if self.reading_order_machine_based: + self.logger.info("Using machine-based detection") + if self.right2left: + self.logger.info("Right-to-left mode enabled") + if self.headers_off: + self.logger.info("Headers ignored in reading order") + if self.reading_order_machine_based: order_text_new, id_of_texts_tot = self.do_order_of_regions_with_model( contours_only_text_parent, contours_only_text_parent_h, text_regions_p) @@ -5108,6 +5299,21 @@ class Eynollah: contours_only_text_parent_d_ordered, contours_only_text_parent_h, boxes_d, textline_mask_tot_d) if self.ocr and self.tr: + self.logger.info("Step 4.5/5: OCR Processing") + + if torch.cuda.is_available(): + self.logger.info("Using GPU acceleration") + else: + self.logger.info("Using CPU processing") + + if self.light_version: + self.logger.info("Using light version OCR") + + if self.textline_light: + self.logger.info("Using light text line detection for OCR") + + self.logger.info("Processing text lines...") + device = cuda.get_current_device() device.reset() gc.collect() @@ -5170,13 +5376,19 @@ class Eynollah: ocr_all_textlines = None ocr_all_textlines_marginals_left = None ocr_all_textlines_marginals_right = None - self.logger.info("detection of reading order took %.1fs", time.time() - t_order) + self.logger.info(f"Detection of reading order took {time.time() - t_order:.1f}s") + + self.logger.info("Step 5/5: Output Generation") + self.logger.info("Generating PAGE-XML output") pcgts = self.writer.build_pagexml_no_full_layout( txt_con_org, page_coord, order_text_new, id_of_texts_tot, all_found_textline_polygons, all_box_coord, polygons_of_images, polygons_of_marginals_left, polygons_of_marginals_right, all_found_textline_polygons_marginals_left, all_found_textline_polygons_marginals_right, all_box_coord_marginals_left, all_box_coord_marginals_right, slopes, slopes_marginals_left, slopes_marginals_right, cont_page, polygons_lines_xml, contours_tables, ocr_all_textlines, ocr_all_textlines_marginals_left, ocr_all_textlines_marginals_right, conf_contours_textregions) + + self.logger.info(f"Output file: {self.writer.output_filename}") + return pcgts @@ -5186,32 +5398,19 @@ class Eynollah_ocr: dir_models, model_name=None, dir_xmls=None, - dir_in=None, - image_filename=None, - dir_in_bin=None, - dir_out=None, - dir_out_image_text=None, tr_ocr=False, batch_size=None, export_textline_images_and_text=False, do_not_mask_with_textline_contour=False, - prediction_with_both_of_rgb_and_bin=False, pref_of_dataset=None, min_conf_value_of_textline_text : Optional[float]=None, logger=None, ): - self.dir_in = dir_in - self.image_filename = image_filename - self.dir_in_bin = dir_in_bin - self.dir_out = dir_out - self.dir_xmls = dir_xmls self.dir_models = dir_models self.model_name = model_name self.tr_ocr = tr_ocr self.export_textline_images_and_text = export_textline_images_and_text self.do_not_mask_with_textline_contour = do_not_mask_with_textline_contour - self.dir_out_image_text = dir_out_image_text - self.prediction_with_both_of_rgb_and_bin = prediction_with_both_of_rgb_and_bin self.pref_of_dataset = pref_of_dataset self.logger = logger if logger else getLogger('eynollah') @@ -5263,24 +5462,27 @@ class Eynollah_ocr: ) self.end_character = len(characters) + 2 - def run(self, overwrite : bool = False): - if self.dir_in: - ls_imgs = os.listdir(self.dir_in) - ls_imgs = [ind_img for ind_img in ls_imgs if ind_img.endswith('.jpg') or ind_img.endswith('.jpeg') or ind_img.endswith('.png') or ind_img.endswith('.tif') or ind_img.endswith('.tiff') or ind_img.endswith('.JPG') or ind_img.endswith('.JPEG') or ind_img.endswith('.TIF') or ind_img.endswith('.TIFF') or ind_img.endswith('.PNG')] + def run(self, overwrite: bool = False, + dir_in: Optional[str] = None, + dir_in_bin: Optional[str] = None, + image_filename: Optional[str] = None, + dir_xmls: Optional[str] = None, + dir_out_image_text: Optional[str] = None, + dir_out: Optional[str] = None, + ): + if dir_in: + ls_imgs = [os.path.join(dir_in, image_filename) + for image_filename in filter(is_image_filename, + os.listdir(dir_in))] else: - ls_imgs = [self.image_filename] - + ls_imgs = [image_filename] + if self.tr_ocr: tr_ocr_input_height_and_width = 384 - for ind_img in ls_imgs: - if self.dir_in: - file_name = Path(ind_img).stem - dir_img = os.path.join(self.dir_in, ind_img) - else: - file_name = Path(self.image_filename).stem - dir_img = self.image_filename - dir_xml = os.path.join(self.dir_xmls, file_name+'.xml') - out_file_ocr = os.path.join(self.dir_out, file_name+'.xml') + for dir_img in ls_imgs: + file_name = Path(dir_img).stem + dir_xml = os.path.join(dir_xmls, file_name+'.xml') + out_file_ocr = os.path.join(dir_out, file_name+'.xml') if os.path.exists(out_file_ocr): if overwrite: @@ -5291,8 +5493,8 @@ class Eynollah_ocr: img = cv2.imread(dir_img) - if self.dir_out_image_text: - out_image_with_text = os.path.join(self.dir_out_image_text, file_name+'.png') + if dir_out_image_text: + out_image_with_text = os.path.join(dir_out_image_text, file_name+'.png') image_text = Image.new("RGB", (img.shape[1], img.shape[0]), "white") draw = ImageDraw.Draw(image_text) total_bb_coordinates = [] @@ -5330,7 +5532,7 @@ class Eynollah_ocr: textline_coords = np.array( [ [ int(x.split(',')[0]) , int(x.split(',')[1]) ] for x in p_h] ) x,y,w,h = cv2.boundingRect(textline_coords) - if self.dir_out_image_text: + if dir_out_image_text: total_bb_coordinates.append([x,y,w,h]) h2w_ratio = h/float(w) @@ -5343,7 +5545,7 @@ class Eynollah_ocr: img_crop = img_poly_on_img[y:y+h, x:x+w, :] img_crop[mask_poly==0] = 255 - + self.logger.debug("processing %d lines for '%s'", len(cropped_lines), nn.attrib['id']) if h2w_ratio > 0.1: cropped_lines.append(resize_image(img_crop, tr_ocr_input_height_and_width, tr_ocr_input_height_and_width) ) cropped_lines_meging_indexing.append(0) @@ -5452,10 +5654,12 @@ class Eynollah_ocr: unique_cropped_lines_region_indexer = np.unique(cropped_lines_region_indexer) - if self.dir_out_image_text: + if dir_out_image_text: - font_path = "Charis-7.000/Charis-Regular.ttf" # Make sure this file exists! - font = ImageFont.truetype(font_path, 40) + #font_path = "Charis-7.000/Charis-Regular.ttf" # Make sure this file exists! + font = importlib_resources.files(__package__) / "Charis-Regular.ttf" + with importlib_resources.as_file(font) as font: + font = ImageFont.truetype(font=font, size=40) for indexer_text, bb_ind in enumerate(total_bb_coordinates): @@ -5465,7 +5669,7 @@ class Eynollah_ocr: w_bb = bb_ind[2] h_bb = bb_ind[3] - font = fit_text_single_line(draw, extracted_texts_merged[indexer_text], font_path, w_bb, int(h_bb*0.4) ) + font = fit_text_single_line(draw, extracted_texts_merged[indexer_text], font.path, w_bb, int(h_bb*0.4) ) ##draw.rectangle([x_bb, y_bb, x_bb + w_bb, y_bb + h_bb], outline="red", width=2) @@ -5580,18 +5784,10 @@ class Eynollah_ocr: img_size=(image_width, image_height) - for ind_img in ls_imgs: - if self.dir_in: - file_name = Path(ind_img).stem - dir_img = os.path.join(self.dir_in, ind_img) - else: - file_name = Path(self.image_filename).stem - dir_img = self.image_filename - - #file_name = Path(ind_img).stem - #dir_img = os.path.join(self.dir_in, ind_img) - dir_xml = os.path.join(self.dir_xmls, file_name+'.xml') - out_file_ocr = os.path.join(self.dir_out, file_name+'.xml') + for dir_img in ls_imgs: + file_name = Path(dir_img).stem + dir_xml = os.path.join(dir_xmls, file_name+'.xml') + out_file_ocr = os.path.join(dir_out, file_name+'.xml') if os.path.exists(out_file_ocr): if overwrite: @@ -5601,13 +5797,13 @@ class Eynollah_ocr: continue img = cv2.imread(dir_img) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: cropped_lines_bin = [] - dir_img_bin = os.path.join(self.dir_in_bin, file_name+'.png') + dir_img_bin = os.path.join(dir_in_bin, file_name+'.png') img_bin = cv2.imread(dir_img_bin) - if self.dir_out_image_text: - out_image_with_text = os.path.join(self.dir_out_image_text, file_name+'.png') + if dir_out_image_text: + out_image_with_text = os.path.join(dir_out_image_text, file_name+'.png') image_text = Image.new("RGB", (img.shape[1], img.shape[0]), "white") draw = ImageDraw.Draw(image_text) total_bb_coordinates = [] @@ -5651,13 +5847,13 @@ class Eynollah_ocr: if type_textregion=='drop-capital': angle_degrees = 0 - if self.dir_out_image_text: + if dir_out_image_text: total_bb_coordinates.append([x,y,w,h]) w_scaled = w * image_height/float(h) img_poly_on_img = np.copy(img) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_poly_on_img_bin = np.copy(img_bin) img_crop_bin = img_poly_on_img_bin[y:y+h, x:x+w, :] @@ -5680,7 +5876,7 @@ class Eynollah_ocr: img_crop = rotate_image_with_padding(img_crop, better_des_slope ) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_crop_bin = rotate_image_with_padding(img_crop_bin, better_des_slope ) mask_poly = rotate_image_with_padding(mask_poly, better_des_slope ) @@ -5695,13 +5891,13 @@ class Eynollah_ocr: if not self.do_not_mask_with_textline_contour: img_crop[mask_poly==0] = 255 - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_crop_bin = img_crop_bin[y_n:y_n+h_n, x_n:x_n+w_n, :] if not self.do_not_mask_with_textline_contour: img_crop_bin[mask_poly==0] = 255 if mask_poly[:,:,0].sum() /float(w_n*h_n) < 0.50 and w_scaled > 90: - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_crop, img_crop_bin = break_curved_line_into_small_pieces_and_then_merge(img_crop, mask_poly, img_crop_bin) else: img_crop, _ = break_curved_line_into_small_pieces_and_then_merge(img_crop, mask_poly) @@ -5711,14 +5907,14 @@ class Eynollah_ocr: better_des_slope = 0 if not self.do_not_mask_with_textline_contour: img_crop[mask_poly==0] = 255 - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: if not self.do_not_mask_with_textline_contour: img_crop_bin[mask_poly==0] = 255 if type_textregion=='drop-capital': pass else: if mask_poly[:,:,0].sum() /float(w*h) < 0.50 and w_scaled > 90: - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_crop, img_crop_bin = break_curved_line_into_small_pieces_and_then_merge(img_crop, mask_poly, img_crop_bin) else: img_crop, _ = break_curved_line_into_small_pieces_and_then_merge(img_crop, mask_poly) @@ -5733,14 +5929,12 @@ class Eynollah_ocr: cropped_lines_ver_index.append(0) cropped_lines_meging_indexing.append(0) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_fin = preprocess_and_resize_image_for_ocrcnn_model(img_crop_bin, image_height, image_width) cropped_lines_bin.append(img_fin) else: - if self.prediction_with_both_of_rgb_and_bin: - splited_images, splited_images_bin = return_textlines_split_if_needed(img_crop, img_crop_bin, prediction_with_both_of_rgb_and_bin=self.prediction_with_both_of_rgb_and_bin) - else: - splited_images, splited_images_bin = return_textlines_split_if_needed(img_crop, None) + splited_images, splited_images_bin = return_textlines_split_if_needed( + img_crop, img_crop_bin if dir_in_bin is not None else None) if splited_images: img_fin = preprocess_and_resize_image_for_ocrcnn_model(splited_images[0], image_height, image_width) cropped_lines.append(img_fin) @@ -5761,7 +5955,7 @@ class Eynollah_ocr: else: cropped_lines_ver_index.append(0) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_fin = preprocess_and_resize_image_for_ocrcnn_model(splited_images_bin[0], image_height, image_width) cropped_lines_bin.append(img_fin) img_fin = preprocess_and_resize_image_for_ocrcnn_model(splited_images_bin[1], image_height, image_width) @@ -5777,7 +5971,7 @@ class Eynollah_ocr: else: cropped_lines_ver_index.append(0) - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: img_fin = preprocess_and_resize_image_for_ocrcnn_model(img_crop_bin, image_height, image_width) cropped_lines_bin.append(img_fin) @@ -5790,29 +5984,15 @@ class Eynollah_ocr: if cheild_text.tag.endswith("Unicode"): textline_text = cheild_text.text if textline_text: - if self.do_not_mask_with_textline_contour: - if self.pref_of_dataset: - with open(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_'+self.pref_of_dataset+'.txt'), 'w') as text_file: - text_file.write(textline_text) - - cv2.imwrite(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_'+self.pref_of_dataset+'.png'), img_crop ) - else: - with open(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'.txt'), 'w') as text_file: - text_file.write(textline_text) - - cv2.imwrite(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'.png'), img_crop ) - else: - if self.pref_of_dataset: - with open(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_'+self.pref_of_dataset+'_masked.txt'), 'w') as text_file: - text_file.write(textline_text) - - cv2.imwrite(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_'+self.pref_of_dataset+'_masked.png'), img_crop ) - else: - with open(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_masked.txt'), 'w') as text_file: - text_file.write(textline_text) - - cv2.imwrite(os.path.join(self.dir_out, file_name+'_line_'+str(indexer_textlines)+'_masked.png'), img_crop ) + base_name = os.path.join(dir_out, file_name + '_line_' + str(indexer_textlines)) + if self.pref_of_dataset: + base_name += '_' + self.pref_of_dataset + if not self.do_not_mask_with_textline_contour: + base_name += '_masked' + with open(base_name + '.txt', 'w') as text_file: + text_file.write(textline_text) + cv2.imwrite(base_name + '.png', img_crop) indexer_textlines+=1 if not self.export_textline_images_and_text: @@ -5843,7 +6023,7 @@ class Eynollah_ocr: else: imgs_ver_flipped = None - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: imgs_bin = cropped_lines_bin[n_start:] imgs_bin = np.array(imgs_bin) imgs_bin = imgs_bin.reshape(imgs_bin.shape[0], image_height, image_width, 3) @@ -5873,7 +6053,7 @@ class Eynollah_ocr: imgs_ver_flipped = None - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: imgs_bin = cropped_lines_bin[n_start:n_end] imgs_bin = np.array(imgs_bin).reshape(self.b_s, image_height, image_width, 3) @@ -5886,6 +6066,7 @@ class Eynollah_ocr: imgs_bin_ver_flipped = None + self.logger.debug("processing next %d lines", len(imgs)) preds = self.prediction_model.predict(imgs, verbose=0) if len(indices_ver)>0: @@ -5912,7 +6093,7 @@ class Eynollah_ocr: if len(indices_where_flipped_conf_value_is_higher)>0: indices_to_be_replaced = indices_ver[indices_where_flipped_conf_value_is_higher] preds[indices_to_be_replaced,:,:] = preds_flipped[indices_where_flipped_conf_value_is_higher, :, :] - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: preds_bin = self.prediction_model.predict(imgs_bin, verbose=0) if len(indices_ver)>0: @@ -5959,7 +6140,7 @@ class Eynollah_ocr: extracted_texts.append("") extracted_conf_value.append(0) del cropped_lines - if self.prediction_with_both_of_rgb_and_bin: + if dir_in_bin is not None: del cropped_lines_bin gc.collect() @@ -5972,10 +6153,12 @@ class Eynollah_ocr: unique_cropped_lines_region_indexer = np.unique(cropped_lines_region_indexer) - if self.dir_out_image_text: + if dir_out_image_text: - font_path = "Charis-7.000/Charis-Regular.ttf" # Make sure this file exists! - font = ImageFont.truetype(font_path, 40) + #font_path = "Charis-7.000/Charis-Regular.ttf" # Make sure this file exists! + font = importlib_resources.files(__package__) / "Charis-Regular.ttf" + with importlib_resources.as_file(font) as font: + font = ImageFont.truetype(font=font, size=40) for indexer_text, bb_ind in enumerate(total_bb_coordinates): @@ -5985,7 +6168,7 @@ class Eynollah_ocr: w_bb = bb_ind[2] h_bb = bb_ind[3] - font = fit_text_single_line(draw, extracted_texts_merged[indexer_text], font_path, w_bb, int(h_bb*0.4) ) + font = fit_text_single_line(draw, extracted_texts_merged[indexer_text], font.path, w_bb, int(h_bb*0.4) ) ##draw.rectangle([x_bb, y_bb, x_bb + w_bb, y_bb + h_bb], outline="red", width=2) diff --git a/src/eynollah/image_enhancer.py b/src/eynollah/image_enhancer.py index 983712d..89dde16 100644 --- a/src/eynollah/image_enhancer.py +++ b/src/eynollah/image_enhancer.py @@ -3,30 +3,24 @@ Image enhancer. The output can be written as same scale of input or in new predi """ from logging import Logger -from difflib import SequenceMatcher as sq -from PIL import Image, ImageDraw, ImageFont -import math import os -import sys import time from typing import Optional import atexit -import warnings from functools import partial from pathlib import Path from multiprocessing import cpu_count import gc -import copy -from loky import ProcessPoolExecutor -import xml.etree.ElementTree as ET import cv2 import numpy as np -from ocrd import OcrdPage from ocrd_utils import getLogger, tf_disable_interactive_logs -import statistics +import tensorflow as tf +from skimage.morphology import skeletonize from tensorflow.keras.models import load_model from .utils.resize import resize_image +from .utils.pil_cv2 import pil2cv from .utils import ( + is_image_filename, crop_image_inside_box ) @@ -38,13 +32,11 @@ class Enhancer: def __init__( self, dir_models : str, - dir_out : Optional[str] = None, num_col_upper : Optional[int] = None, num_col_lower : Optional[int] = None, save_org_scale : bool = False, logger : Optional[Logger] = None, ): - self.dir_out = dir_out self.input_binary = False self.light_version = False self.save_org_scale = save_org_scale @@ -58,13 +50,11 @@ class Enhancer: self.num_col_lower = num_col_lower self.logger = logger if logger else getLogger('enhancement') - # for parallelization of CPU-intensive tasks: - self.executor = ProcessPoolExecutor(max_workers=cpu_count(), timeout=1200) - atexit.register(self.executor.shutdown) self.dir_models = dir_models + self.model_dir_of_binarization = dir_models + "/eynollah-binarization_20210425" self.model_dir_of_enhancement = dir_models + "/eynollah-enhancement_20210425" self.model_dir_of_col_classifier = dir_models + "/eynollah-column-classifier_20210425" - self.model_page_dir = dir_models + "/eynollah-page-extraction_20210425" + self.model_page_dir = dir_models + "/model_eynollah_page_extraction_20250915" try: for device in tf.config.list_physical_devices('GPU'): @@ -75,10 +65,10 @@ class Enhancer: self.model_page = self.our_load_model(self.model_page_dir) self.model_classifier = self.our_load_model(self.model_dir_of_col_classifier) self.model_enhancement = self.our_load_model(self.model_dir_of_enhancement) + self.model_bin = self.our_load_model(self.model_dir_of_binarization) def cache_images(self, image_filename=None, image_pil=None, dpi=None): ret = {} - t_c0 = time.time() if image_filename: ret['img'] = cv2.imread(image_filename) if self.light_version: @@ -98,10 +88,9 @@ class Enhancer: if dpi is not None: self.dpi = dpi - def reset_file_name_dir(self, image_filename): - t_c = time.time() + def reset_file_name_dir(self, image_filename, dir_out): self.cache_images(image_filename=image_filename) - self.output_filename = os.path.join(self.dir_out, Path(image_filename).stem +'.png') + self.output_filename = os.path.join(dir_out, Path(image_filename).stem +'.png') def imread(self, grayscale=False, uint8=True): key = 'img' @@ -699,7 +688,12 @@ class Enhancer: return img_res - def run(self, image_filename : Optional[str] = None, dir_in : Optional[str] = None, overwrite : bool = False): + def run(self, + overwrite: bool = False, + image_filename: Optional[str] = None, + dir_in: Optional[str] = None, + dir_out: Optional[str] = None, + ): """ Get image and scales, then extract the page of scanned image """ @@ -707,17 +701,19 @@ class Enhancer: t0_tot = time.time() if dir_in: - self.ls_imgs = os.listdir(dir_in) + ls_imgs = [os.path.join(dir_in, image_filename) + for image_filename in filter(is_image_filename, + os.listdir(dir_in))] elif image_filename: - self.ls_imgs = [image_filename] + ls_imgs = [image_filename] else: raise ValueError("run requires either a single image filename or a directory") - for img_filename in self.ls_imgs: + for img_filename in ls_imgs: self.logger.info(img_filename) t0 = time.time() - self.reset_file_name_dir(os.path.join(dir_in or "", img_filename)) + self.reset_file_name_dir(img_filename, dir_out) #print("text region early -11 in %.1fs", time.time() - t0) if os.path.exists(self.output_filename): diff --git a/src/eynollah/mb_ro_on_layout.py b/src/eynollah/mb_ro_on_layout.py index c03d831..45db8e4 100644 --- a/src/eynollah/mb_ro_on_layout.py +++ b/src/eynollah/mb_ro_on_layout.py @@ -3,48 +3,28 @@ Image enhancer. The output can be written as same scale of input or in new predi """ from logging import Logger -from difflib import SequenceMatcher as sq -from PIL import Image, ImageDraw, ImageFont -import math import os -import sys import time from typing import Optional import atexit -import warnings from functools import partial from pathlib import Path from multiprocessing import cpu_count -import gc -import copy -from loky import ProcessPoolExecutor import xml.etree.ElementTree as ET import cv2 import numpy as np -from ocrd import OcrdPage -from ocrd_utils import getLogger, tf_disable_interactive_logs +from ocrd_utils import getLogger import statistics +import tensorflow as tf from tensorflow.keras.models import load_model from .utils.resize import resize_image -from .utils import ( - crop_image_inside_box -) from .utils.contour import ( - filter_contours_area_of_image, - filter_contours_area_of_image_tables, - find_contours_mean_y_diff, find_new_features_of_contours, - find_features_of_contours, - get_text_region_boxes_by_given_contours, - get_textregion_contours_in_org_image, - get_textregion_contours_in_org_image_light, return_contours_of_image, - return_contours_of_interested_region, - return_contours_of_interested_region_by_min_size, - return_contours_of_interested_textline, return_parent_contours, ) +from .utils import is_xml_filename DPI_THRESHOLD = 298 KERNEL = np.ones((5, 5), np.uint8) @@ -54,17 +34,11 @@ class machine_based_reading_order_on_layout: def __init__( self, dir_models : str, - dir_out : Optional[str] = None, logger : Optional[Logger] = None, ): - self.dir_out = dir_out - - self.logger = logger if logger else getLogger('mbro on layout') - # for parallelization of CPU-intensive tasks: - self.executor = ProcessPoolExecutor(max_workers=cpu_count(), timeout=1200) - atexit.register(self.executor.shutdown) + self.logger = logger if logger else getLogger('mbreorder') self.dir_models = dir_models - self.model_reading_order_dir = dir_models + "/model_eynollah_reading_order_20250824"#"/model_ens_reading_order_machine_based" + self.model_reading_order_dir = dir_models + "/model_eynollah_reading_order_20250824" try: for device in tf.config.list_physical_devices('GPU'): @@ -75,45 +49,6 @@ class machine_based_reading_order_on_layout: self.model_reading_order = self.our_load_model(self.model_reading_order_dir) self.light_version = True - - def cache_images(self, image_filename=None, image_pil=None, dpi=None): - ret = {} - t_c0 = time.time() - if image_filename: - ret['img'] = cv2.imread(image_filename) - if self.light_version: - self.dpi = 100 - else: - self.dpi = 0#check_dpi(image_filename) - else: - ret['img'] = pil2cv(image_pil) - if self.light_version: - self.dpi = 100 - else: - self.dpi = 0#check_dpi(image_pil) - ret['img_grayscale'] = cv2.cvtColor(ret['img'], cv2.COLOR_BGR2GRAY) - for prefix in ('', '_grayscale'): - ret[f'img{prefix}_uint8'] = ret[f'img{prefix}'].astype(np.uint8) - self._imgs = ret - if dpi is not None: - self.dpi = dpi - - def reset_file_name_dir(self, image_filename): - t_c = time.time() - self.cache_images(image_filename=image_filename) - self.output_filename = os.path.join(self.dir_out, Path(image_filename).stem +'.png') - - def imread(self, grayscale=False, uint8=True): - key = 'img' - if grayscale: - key += '_grayscale' - if uint8: - key += '_uint8' - return self._imgs[key].copy() - - def isNaN(self, num): - return num != num - @staticmethod def our_load_model(model_file): if model_file.endswith('.h5') and Path(model_file[:-3]).exists(): @@ -125,281 +60,8 @@ class machine_based_reading_order_on_layout: model = load_model(model_file, compile=False, custom_objects={ "PatchEncoder": PatchEncoder, "Patches": Patches}) return model - - def predict_enhancement(self, img): - self.logger.debug("enter predict_enhancement") - img_height_model = self.model_enhancement.layers[-1].output_shape[1] - img_width_model = self.model_enhancement.layers[-1].output_shape[2] - if img.shape[0] < img_height_model: - img = cv2.resize(img, (img.shape[1], img_width_model), interpolation=cv2.INTER_NEAREST) - if img.shape[1] < img_width_model: - img = cv2.resize(img, (img_height_model, img.shape[0]), interpolation=cv2.INTER_NEAREST) - margin = int(0.1 * img_width_model) - width_mid = img_width_model - 2 * margin - height_mid = img_height_model - 2 * margin - img = img / 255. - img_h = img.shape[0] - img_w = img.shape[1] - - prediction_true = np.zeros((img_h, img_w, 3)) - nxf = img_w / float(width_mid) - nyf = img_h / float(height_mid) - nxf = int(nxf) + 1 if nxf > int(nxf) else int(nxf) - nyf = int(nyf) + 1 if nyf > int(nyf) else int(nyf) - - for i in range(nxf): - for j in range(nyf): - if i == 0: - index_x_d = i * width_mid - index_x_u = index_x_d + img_width_model - else: - index_x_d = i * width_mid - index_x_u = index_x_d + img_width_model - if j == 0: - index_y_d = j * height_mid - index_y_u = index_y_d + img_height_model - else: - index_y_d = j * height_mid - index_y_u = index_y_d + img_height_model - - if index_x_u > img_w: - index_x_u = img_w - index_x_d = img_w - img_width_model - if index_y_u > img_h: - index_y_u = img_h - index_y_d = img_h - img_height_model - - img_patch = img[np.newaxis, index_y_d:index_y_u, index_x_d:index_x_u, :] - label_p_pred = self.model_enhancement.predict(img_patch, verbose=0) - seg = label_p_pred[0, :, :, :] * 255 - - if i == 0 and j == 0: - prediction_true[index_y_d + 0:index_y_u - margin, - index_x_d + 0:index_x_u - margin] = \ - seg[0:-margin or None, - 0:-margin or None] - elif i == nxf - 1 and j == nyf - 1: - prediction_true[index_y_d + margin:index_y_u - 0, - index_x_d + margin:index_x_u - 0] = \ - seg[margin:, - margin:] - elif i == 0 and j == nyf - 1: - prediction_true[index_y_d + margin:index_y_u - 0, - index_x_d + 0:index_x_u - margin] = \ - seg[margin:, - 0:-margin or None] - elif i == nxf - 1 and j == 0: - prediction_true[index_y_d + 0:index_y_u - margin, - index_x_d + margin:index_x_u - 0] = \ - seg[0:-margin or None, - margin:] - elif i == 0 and j != 0 and j != nyf - 1: - prediction_true[index_y_d + margin:index_y_u - margin, - index_x_d + 0:index_x_u - margin] = \ - seg[margin:-margin or None, - 0:-margin or None] - elif i == nxf - 1 and j != 0 and j != nyf - 1: - prediction_true[index_y_d + margin:index_y_u - margin, - index_x_d + margin:index_x_u - 0] = \ - seg[margin:-margin or None, - margin:] - elif i != 0 and i != nxf - 1 and j == 0: - prediction_true[index_y_d + 0:index_y_u - margin, - index_x_d + margin:index_x_u - margin] = \ - seg[0:-margin or None, - margin:-margin or None] - elif i != 0 and i != nxf - 1 and j == nyf - 1: - prediction_true[index_y_d + margin:index_y_u - 0, - index_x_d + margin:index_x_u - margin] = \ - seg[margin:, - margin:-margin or None] - else: - prediction_true[index_y_d + margin:index_y_u - margin, - index_x_d + margin:index_x_u - margin] = \ - seg[margin:-margin or None, - margin:-margin or None] - - prediction_true = prediction_true.astype(int) - return prediction_true - - def calculate_width_height_by_columns(self, img, num_col, width_early, label_p_pred): - self.logger.debug("enter calculate_width_height_by_columns") - if num_col == 1: - img_w_new = 2000 - elif num_col == 2: - img_w_new = 2400 - elif num_col == 3: - img_w_new = 3000 - elif num_col == 4: - img_w_new = 4000 - elif num_col == 5: - img_w_new = 5000 - elif num_col == 6: - img_w_new = 6500 - else: - img_w_new = width_early - img_h_new = img_w_new * img.shape[0] // img.shape[1] - - if img_h_new >= 8000: - img_new = np.copy(img) - num_column_is_classified = False - else: - img_new = resize_image(img, img_h_new, img_w_new) - num_column_is_classified = True - - return img_new, num_column_is_classified - - def early_page_for_num_of_column_classification(self,img_bin): - self.logger.debug("enter early_page_for_num_of_column_classification") - if self.input_binary: - img = np.copy(img_bin).astype(np.uint8) - else: - img = self.imread() - img = cv2.GaussianBlur(img, (5, 5), 0) - img_page_prediction = self.do_prediction(False, img, self.model_page) - - imgray = cv2.cvtColor(img_page_prediction, cv2.COLOR_BGR2GRAY) - _, thresh = cv2.threshold(imgray, 0, 255, 0) - thresh = cv2.dilate(thresh, KERNEL, iterations=3) - contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) - if len(contours)>0: - cnt_size = np.array([cv2.contourArea(contours[j]) - for j in range(len(contours))]) - cnt = contours[np.argmax(cnt_size)] - box = cv2.boundingRect(cnt) - else: - box = [0, 0, img.shape[1], img.shape[0]] - cropped_page, page_coord = crop_image_inside_box(box, img) - - self.logger.debug("exit early_page_for_num_of_column_classification") - return cropped_page, page_coord - - def calculate_width_height_by_columns_1_2(self, img, num_col, width_early, label_p_pred): - self.logger.debug("enter calculate_width_height_by_columns") - if num_col == 1: - img_w_new = 1000 - else: - img_w_new = 1300 - img_h_new = img_w_new * img.shape[0] // img.shape[1] - - if label_p_pred[0][int(num_col - 1)] < 0.9 and img_w_new < width_early: - img_new = np.copy(img) - num_column_is_classified = False - #elif label_p_pred[0][int(num_col - 1)] < 0.8 and img_h_new >= 8000: - elif img_h_new >= 8000: - img_new = np.copy(img) - num_column_is_classified = False - else: - img_new = resize_image(img, img_h_new, img_w_new) - num_column_is_classified = True - - return img_new, num_column_is_classified - - def resize_and_enhance_image_with_column_classifier(self, light_version): - self.logger.debug("enter resize_and_enhance_image_with_column_classifier") - dpi = 0#self.dpi - self.logger.info("Detected %s DPI", dpi) - if self.input_binary: - img = self.imread() - prediction_bin = self.do_prediction(True, img, self.model_bin, n_batch_inference=5) - prediction_bin = 255 * (prediction_bin[:,:,0]==0) - prediction_bin = np.repeat(prediction_bin[:, :, np.newaxis], 3, axis=2).astype(np.uint8) - img= np.copy(prediction_bin) - img_bin = prediction_bin - else: - img = self.imread() - self.h_org, self.w_org = img.shape[:2] - img_bin = None - - width_early = img.shape[1] - t1 = time.time() - _, page_coord = self.early_page_for_num_of_column_classification(img_bin) - - self.image_page_org_size = img[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3], :] - self.page_coord = page_coord - - if self.num_col_upper and not self.num_col_lower: - num_col = self.num_col_upper - label_p_pred = [np.ones(6)] - elif self.num_col_lower and not self.num_col_upper: - num_col = self.num_col_lower - label_p_pred = [np.ones(6)] - elif not self.num_col_upper and not self.num_col_lower: - if self.input_binary: - img_in = np.copy(img) - img_in = img_in / 255.0 - img_in = cv2.resize(img_in, (448, 448), interpolation=cv2.INTER_NEAREST) - img_in = img_in.reshape(1, 448, 448, 3) - else: - img_1ch = self.imread(grayscale=True) - width_early = img_1ch.shape[1] - img_1ch = img_1ch[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3]] - - img_1ch = img_1ch / 255.0 - img_1ch = cv2.resize(img_1ch, (448, 448), interpolation=cv2.INTER_NEAREST) - img_in = np.zeros((1, img_1ch.shape[0], img_1ch.shape[1], 3)) - img_in[0, :, :, 0] = img_1ch[:, :] - img_in[0, :, :, 1] = img_1ch[:, :] - img_in[0, :, :, 2] = img_1ch[:, :] - - label_p_pred = self.model_classifier.predict(img_in, verbose=0) - num_col = np.argmax(label_p_pred[0]) + 1 - elif (self.num_col_upper and self.num_col_lower) and (self.num_col_upper!=self.num_col_lower): - if self.input_binary: - img_in = np.copy(img) - img_in = img_in / 255.0 - img_in = cv2.resize(img_in, (448, 448), interpolation=cv2.INTER_NEAREST) - img_in = img_in.reshape(1, 448, 448, 3) - else: - img_1ch = self.imread(grayscale=True) - width_early = img_1ch.shape[1] - img_1ch = img_1ch[page_coord[0] : page_coord[1], page_coord[2] : page_coord[3]] - - img_1ch = img_1ch / 255.0 - img_1ch = cv2.resize(img_1ch, (448, 448), interpolation=cv2.INTER_NEAREST) - img_in = np.zeros((1, img_1ch.shape[0], img_1ch.shape[1], 3)) - img_in[0, :, :, 0] = img_1ch[:, :] - img_in[0, :, :, 1] = img_1ch[:, :] - img_in[0, :, :, 2] = img_1ch[:, :] - - label_p_pred = self.model_classifier.predict(img_in, verbose=0) - num_col = np.argmax(label_p_pred[0]) + 1 - - if num_col > self.num_col_upper: - num_col = self.num_col_upper - label_p_pred = [np.ones(6)] - if num_col < self.num_col_lower: - num_col = self.num_col_lower - label_p_pred = [np.ones(6)] - else: - num_col = self.num_col_upper - label_p_pred = [np.ones(6)] - - self.logger.info("Found %d columns (%s)", num_col, np.around(label_p_pred, decimals=5)) - - if dpi < DPI_THRESHOLD: - if light_version and num_col in (1,2): - img_new, num_column_is_classified = self.calculate_width_height_by_columns_1_2( - img, num_col, width_early, label_p_pred) - else: - img_new, num_column_is_classified = self.calculate_width_height_by_columns( - img, num_col, width_early, label_p_pred) - if light_version: - image_res = np.copy(img_new) - else: - image_res = self.predict_enhancement(img_new) - is_image_enhanced = True - - else: - num_column_is_classified = True - image_res = np.copy(img) - is_image_enhanced = False - - self.logger.debug("exit resize_and_enhance_image_with_column_classifier") - return is_image_enhanced, img, image_res, num_col, num_column_is_classified, img_bin def read_xml(self, xml_file): - file_name = Path(xml_file).stem tree1 = ET.parse(xml_file, parser = ET.XMLParser(encoding='utf-8')) root1=tree1.getroot() alltags=[elem.tag for elem in root1.iter()] @@ -821,7 +483,7 @@ class machine_based_reading_order_on_layout: img_poly=cv2.fillPoly(img, pts =co_img, color=(4,4,4)) img_poly=cv2.fillPoly(img, pts =co_sep, color=(5,5,5)) - return tree1, root1, bb_coord_printspace, file_name, id_paragraph, id_header+id_heading, co_text_paragraph, co_text_header+co_text_heading,\ + return tree1, root1, bb_coord_printspace, id_paragraph, id_header+id_heading, co_text_paragraph, co_text_header+co_text_heading,\ tot_region_ref,x_len, y_len,index_tot_regions, img_poly def return_indexes_of_contours_loctaed_inside_another_list_of_contours(self, contours, contours_loc, cx_main_loc, cy_main_loc, indexes_loc): @@ -1070,7 +732,12 @@ class machine_based_reading_order_on_layout: - def run(self, xml_filename : Optional[str] = None, dir_in : Optional[str] = None, overwrite : bool = False): + def run(self, + overwrite: bool = False, + xml_filename: Optional[str] = None, + dir_in: Optional[str] = None, + dir_out: Optional[str] = None, + ): """ Get image and scales, then extract the page of scanned image """ @@ -1078,22 +745,22 @@ class machine_based_reading_order_on_layout: t0_tot = time.time() if dir_in: - self.ls_xmls = os.listdir(dir_in) + ls_xmls = [os.path.join(dir_in, xml_filename) + for xml_filename in filter(is_xml_filename, + os.listdir(dir_in))] elif xml_filename: - self.ls_xmls = [xml_filename] + ls_xmls = [xml_filename] else: raise ValueError("run requires either a single image filename or a directory") - for xml_filename in self.ls_xmls: + for xml_filename in ls_xmls: self.logger.info(xml_filename) t0 = time.time() - - if dir_in: - xml_file = os.path.join(dir_in, xml_filename) - else: - xml_file = xml_filename - - tree_xml, root_xml, bb_coord_printspace, file_name, id_paragraph, id_header, co_text_paragraph, co_text_header, tot_region_ref, x_len, y_len, index_tot_regions, img_poly = self.read_xml(xml_file) + + file_name = Path(xml_filename).stem + (tree_xml, root_xml, bb_coord_printspace, id_paragraph, id_header, + co_text_paragraph, co_text_header, tot_region_ref, + x_len, y_len, index_tot_regions, img_poly) = self.read_xml(xml_filename) id_all_text = id_paragraph + id_header @@ -1136,7 +803,11 @@ class machine_based_reading_order_on_layout: alltags=[elem.tag for elem in root_xml.iter()] ET.register_namespace("",name_space) - tree_xml.write(os.path.join(self.dir_out, file_name+'.xml'),xml_declaration=True,method='xml',encoding="utf8",default_namespace=None) + tree_xml.write(os.path.join(dir_out, file_name+'.xml'), + xml_declaration=True, + method='xml', + encoding="utf8", + default_namespace=None) #sys.exit() diff --git a/src/eynollah/ocrd-tool.json b/src/eynollah/ocrd-tool.json index af5e03f..5d89c92 100644 --- a/src/eynollah/ocrd-tool.json +++ b/src/eynollah/ocrd-tool.json @@ -1,5 +1,5 @@ { - "version": "0.4.0", + "version": "0.5.0", "git_url": "https://github.com/qurator-spk/eynollah", "dockerhub": "ocrd/eynollah", "tools": { @@ -82,13 +82,23 @@ } }, "resources": [ + { + "url": "https://zenodo.org/records/17194824/files/models_layout_v0_5_0.tar.gz?download=1", + "name": "models_layout_v0_5_0", + "type": "archive", + "path_in_archive": "models_layout_v0_5_0", + "size": 3525684179, + "description": "Models for layout detection, reading order detection, textline detection, page extraction, column classification, table detection, binarization, image enhancement", + "version_range": ">= v0.5.0" + }, { "description": "models for eynollah (TensorFlow SavedModel format)", "url": "https://github.com/qurator-spk/eynollah/releases/download/v0.3.1/models_eynollah.tar.gz", "name": "default", "size": 1894627041, "type": "archive", - "path_in_archive": "models_eynollah" + "path_in_archive": "models_eynollah", + "version_range": ">= v0.3.0, < v0.5.0" } ] }, diff --git a/src/eynollah/processor.py b/src/eynollah/processor.py index c2922c1..12c7356 100644 --- a/src/eynollah/processor.py +++ b/src/eynollah/processor.py @@ -1,6 +1,7 @@ +from functools import cached_property from typing import Optional from ocrd_models import OcrdPage -from ocrd import Processor, OcrdPageResult +from ocrd import OcrdPageResultImage, Processor, OcrdPageResult from .eynollah import Eynollah, EynollahXmlWriter @@ -9,8 +10,8 @@ class EynollahProcessor(Processor): # already employs GPU (without singleton process atm) max_workers = 1 - @property - def executable(self): + @cached_property + def executable(self) -> str: return 'ocrd-eynollah-segment' def setup(self) -> None: @@ -20,7 +21,6 @@ class EynollahProcessor(Processor): "and parameter 'light_version' (faster+simpler method for main region detection and deskewing)") self.eynollah = Eynollah( self.resolve_resource(self.parameter['models']), - logger=self.logger, allow_enhancement=self.parameter['allow_enhancement'], curved_line=self.parameter['curved_line'], right2left=self.parameter['right_to_left'], @@ -33,6 +33,7 @@ class EynollahProcessor(Processor): headers_off=self.parameter['headers_off'], tables=self.parameter['tables'], ) + self.eynollah.logger = self.logger self.eynollah.plotter = None def shutdown(self): diff --git a/src/eynollah/sbb_binarize.py b/src/eynollah/sbb_binarize.py index f43b6ba..3716987 100644 --- a/src/eynollah/sbb_binarize.py +++ b/src/eynollah/sbb_binarize.py @@ -16,6 +16,7 @@ import tensorflow as tf from tensorflow.keras.models import load_model from tensorflow.python.keras import backend as tensorflow_backend +from .utils import is_image_filename def resize_image(img_in, input_height, input_width): return cv2.resize(img_in, (input_width, input_height), interpolation=cv2.INTER_NEAREST) @@ -314,8 +315,8 @@ class SbbBinarizer: prediction_true = prediction_true.astype(np.uint8) return prediction_true[:,:,0] - def run(self, image=None, image_path=None, save=None, use_patches=False, dir_in=None, dir_out=None): - print(dir_in,'dir_in') + def run(self, image=None, image_path=None, output=None, use_patches=False, dir_in=None): + # print(dir_in,'dir_in') if not dir_in: if (image is not None and image_path is not None) or \ (image is None and image_path is None): @@ -343,11 +344,11 @@ class SbbBinarizer: kernel = np.ones((5, 5), np.uint8) img_last[:, :][img_last[:, :] > 0] = 255 img_last = (img_last[:, :] == 0) * 255 - if save: - cv2.imwrite(save, img_last) + if output: + cv2.imwrite(output, img_last) return img_last else: - ls_imgs = os.listdir(dir_in) + ls_imgs = list(filter(is_image_filename, os.listdir(dir_in))) for image_name in ls_imgs: image_stem = image_name.split('.')[0] print(image_name,'image_name') @@ -374,4 +375,4 @@ class SbbBinarizer: img_last[:, :][img_last[:, :] > 0] = 255 img_last = (img_last[:, :] == 0) * 255 - cv2.imwrite(os.path.join(dir_out,image_stem+'.png'), img_last) + cv2.imwrite(os.path.join(output, image_stem + '.png'), img_last) diff --git a/src/eynollah/utils/__init__.py b/src/eynollah/utils/__init__.py index ca86047..6eeabd0 100644 --- a/src/eynollah/utils/__init__.py +++ b/src/eynollah/utils/__init__.py @@ -2194,3 +2194,14 @@ def return_boxes_of_images_by_order_of_reading_new( return boxes, peaks_neg_tot_tables_new else: return boxes, peaks_neg_tot_tables + +def is_image_filename(fname: str) -> bool: + return fname.lower().endswith(('.jpg', + '.jpeg', + '.png', + '.tif', + '.tiff', + )) + +def is_xml_filename(fname: str) -> bool: + return fname.lower().endswith('.xml') diff --git a/src/eynollah/utils/utils_ocr.py b/src/eynollah/utils/utils_ocr.py index d974650..4fa99f7 100644 --- a/src/eynollah/utils/utils_ocr.py +++ b/src/eynollah/utils/utils_ocr.py @@ -109,13 +109,13 @@ def fit_text_single_line(draw, text, font_path, max_width, max_height): return ImageFont.truetype(font_path, 10) # Smallest font fallback -def return_textlines_split_if_needed(textline_image, textline_image_bin, prediction_with_both_of_rgb_and_bin=False): +def return_textlines_split_if_needed(textline_image, textline_image_bin=None): split_point = return_start_and_end_of_common_text_of_textline_ocr_without_common_section(textline_image) if split_point: image1 = textline_image[:, :split_point,:]# image.crop((0, 0, width2, height)) image2 = textline_image[:, split_point:,:]#image.crop((width1, 0, width, height)) - if prediction_with_both_of_rgb_and_bin: + if textline_image_bin is not None: image1_bin = textline_image_bin[:, :split_point,:]# image.crop((0, 0, width2, height)) image2_bin = textline_image_bin[:, split_point:,:]#image.crop((width1, 0, width, height)) return [image1, image2], [image1_bin, image2_bin] diff --git a/tests/resources/euler_rechenkunst01_1738_0025.xml b/tests/resources/euler_rechenkunst01_1738_0025.xml new file mode 100644 index 0000000..1a92f73 --- /dev/null +++ b/tests/resources/euler_rechenkunst01_1738_0025.xml @@ -0,0 +1,1626 @@ + + + OCR-D + 2016-09-29T14:32:09 + 2018-04-25T08:56:33 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + 9 + + + 9 + + + 9 + + + + + + + + + der + + + + + rechten + + + + + gegen + + + + + der + + + + + lincken + + + + + Hand + + + + + bedeutet + + + der rechten gegen der lincken Hand bedeutet + + + + + + + + wie + + + + + folget: + + + wie folget: + + + der rechten gegen der lincken Hand bedeutet +wie folget: + + + + + + + + + I. + + + I. + + + I. + + + + + + + + + 0 + + + + + - + + + + + nichts + + + 0 - nichts + + + + + + + + 1 + + + + + - + + + + + eins + + + 1 - eins + + + + + + + + 2 + + + + + - + + + + + zwey + + + 2 - zwey + + + + + + + + 3 + + + + + - + + + + + drey + + + 3 - drey + + + + + + + + 4 + + + + + - + + + + + vier + + + 4 - vier + + + 0 - nichts +1 - eins +2 - zwey +3 - drey +4 - vier + + + + + + + + + 5 + + + + + - + + + + + fuͤnf + + + 5 - fuͤnf + + + + + + + + 6 + + + + + - + + + + + ſechs + + + 6 - ſechs + + + + + + + 7 + + + + + - + + + + + ſieben + + + 7 - ſieben + + + + + + + + 8 + + + + + - + + + + + acht + + + 8 - acht + + + + + + + + 9 + + + + + - + + + + + neun + + + 9 - neun + + + 5 - fuͤnf +6 - ſechs +7 - ſieben +8 - acht +9 - neun + + + + + + + + + Auf + + + + + der + + + + + zweyten + + + + + Stelle + + + + + aber + + + + + bedeutet. + + + Auf der zweyten Stelle aber bedeutet. + + + Auf der zweyten Stelle aber bedeutet. + + + + + + + + + II. + + + II. + + + II. + + + + + + + + + 0 + + + + + - + + + + + nichts + + + 0 - nichts + + + + + + + + 1 + + + + + - + + + + + zehen + + + 1 - zehen + + + + + + + + 2 + + + + + - + + + + + zwanzig + + + 2 - zwanzig + + + + + + + 3 + + + + + - + + + + + dreyßig + + + 3 - dreyßig + + + + + + + 4 + + + + + - + + + + + vierzig + + + 4 - vierzig + + 0 - nichts +1 - zehen +2 - zwanzig +3 - dreyßig +4 - vierzig + + + + + + + + + 5 + + + + + - + + + + + fuͤnfzig + + + 5 - fuͤnfzig + + + + + + + + 6 + + + + + - + + + + + ſechzig + + + 6 - ſechzig + + + + + + + 7 + + + + + - + + + + + ſiebenzig + + + 7 - ſiebenzig + + + + + + + 8 + + + + + - + + + + + achtzig + + + 8 - achtzig + + + + + + + 9 + + + + + - + + + + + neunzig + + + 9 - neunzig + + 5 - fuͤnfzig +6 - ſechzig +7 - ſiebenzig +8 - achtzig +9 - neunzig + + + + + + + + + Auf + + + + + der + + + + + dritten + + + + + Stelle + + + + + bedeutet. + + + Auf der dritten Stelle bedeutet. + + + Auf der dritten Stelle bedeutet. + + + + + + + + + III. + + + III. + + + III. + + + + + + + + + 0 + + + + + - + + + + + nichts + + + 0 - nichts + + + + + + + + 1 + + + + + - + + + + + hundert + + + 1 - hundert + + + + + + + + 2 + + + + + - + + + + + zwey + + + + + hundert + + + 2 - zwey hundert + + + + + + + + 3 + + + + + - + + + + + drey + + + + + hundert + + + 3 - drey hundert + + + + + + + + 4 + + + + + - + + + + + vier + + + + + hundert + + + 4 - vier hundert + + + 0 - nichts +1 - hundert +2 - zwey hundert +3 - drey hundert +4 - vier hundert + + + + + + + + + 5 + + + + + - + + + + + fuͤnf + + + + + hundert + + + 5 - fuͤnf hundert + + + + + + + + 6 + + + + + - + + + + + ſechs + + + + + hundert + + + 6 - ſechs hundert + + + + + + + 7 + + + + + - + + + + + ſieben + + + + + hundert + + + 7 - ſieben hundert + + + + + + + + 8 + + + + + - + + + + + acht + + + + + hundert + + + 8 - acht hundert + + + + + + + 9 + + + + + - + + + + + neun + + + + + hundert + + + 9 - neun hundert + + + 5 - fuͤnf hundert +6 - ſechs hundert +7 - ſieben hundert +8 - acht hundert +9 - neun hundert + + + + + + + + + Auf + + + + + der + + + + + vierten + + + + + Stelle + + + + + bedeutet. + + + Auf der vierten Stelle bedeutet. + + + Auf der vierten Stelle bedeutet. + + + + + + + + + IV. + + + IV. + + + IV. + + + + + + + + + 0 + + + + + - + + + + + nichts + + + 0 - nichts + + + + + + + + 1 + + + + + - + + + + + tauſend + + + 1 - tauſend + + + + + + + + 2 + + + + + - + + + + + zwey + + + + + tauſend + + + 2 - zwey tauſend + + + + + + + + 3 + + + + + - + + + + + drey + + + + + tauſend + + + 3 - drey tauſend + + + + + + + + 4 + + + + + - + + + + + vier + + + + + tauſend + + + 4 - vier tauſend + + + 0 - nichts +1 - tauſend +2 - zwey tauſend +3 - drey tauſend +4 - vier tauſend + + + + + + + + + 5 + + + + + - + + + + + fuͤnf + + + + + tauſend + + + 5 - fuͤnf tauſend + + + + + + + + 6 + + + + + - + + + + + ſechs + + + + + tauſend + + + 6 - ſechs tauſend + + + + + + + 7 + + + + + - + + + + + ſieben + + + + + tauſend + + + 7 - ſieben tauſend + + + + + + + + 8 + + + + + - + + + + + acht + + + + + tauſend + + + 8 - acht tauſend + + + + + + + 9 + + + + + - + + + + + neun + + + + + tauſend + + + 9 - neun tauſend + + 5 - fuͤnf tauſend +6 - ſechs tauſend +7 - ſieben tauſend +8 - acht tauſend +9 - neun tauſend + + + + + + + + + Auf + + + + + der + + + + + fuͤnften + + + + + Stelle + + + + + bedeutet. + + + Auf der fuͤnften Stelle bedeutet. + + + Auf der fuͤnften Stelle bedeutet. + + + + + + + + + V. + + + V. + + + V. + + + + + + + + + 0 + + + + + - + + + + + nichts + + + 0 - nichts + + + + + + + + 1 + + + + + - + + + + + zehen + + + + + tauſend + + + 1 - zehen tauſend + + + + + + + + 2 + + + + + - + + + + + zwanzig + + + + + tauſend + + + 2 - zwanzig tauſend + + + + + + + + 3 + + + + + - + + + + + dreyßig + + + + + tauſend + + + 3 - dreyßig tauſend + + + + + + + + 4 + + + + + - + + + + + vierzig + + + + + tauſend + + + 4 - vierzig tauſend + + + 0 - nichts +1 - zehen tauſend +2 - zwanzig tauſend +3 - dreyßig tauſend +4 - vierzig tauſend + + + + + + + + + 5 + + + + + - + + + + + fuͤnfzig + + + + + tauſend + + + 5 - fuͤnfzig tauſend + + + + + + + + 6 + + + + + - + + + + + ſechzig + + + + + tauſend + + + 6 - ſechzig tauſend + + + + + + + 7 + + + + + - + + + + + ſiebenzig + + + + + tauſend + + + 7 - ſiebenzig tauſend + + + + + + + + 8 + + + + + - + + + + + achtzig + + + + + tauſend + + + 8 - achtzig tauſend + + + + + + + 9 + + + + + - + + + + + neunzig + + + + + tauſend + + + 9 - neunzig tauſend + + + 5 - fuͤnfzig tauſend +6 - ſechzig tauſend +7 - ſiebenzig tauſend +8 - achtzig tauſend +9 - neunzig tauſend + + + + + + + + A + + + + + 5 + + + A 5 + + A 5 + + + + + + + + + Anf + + + Anf + + Anf + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/tests/resources/kant_aufklaerung_1784_0020.xml b/tests/resources/kant_aufklaerung_1784_0020.xml new file mode 100644 index 0000000..47484cd --- /dev/null +++ b/tests/resources/kant_aufklaerung_1784_0020.xml @@ -0,0 +1,2129 @@ + + + OCR-D + 2016-09-20T11:09:27.431+02:00 + 2018-04-24T17:44:49.605+01:00 + + + + + + + + + + + + + + + + + + + + + + + ( + + + + + + + 484 + + + + + + + ) + + + + + ( 484 ) + + + + ( 484 ) + + + + + + + + + + + gewiegelt + + + + + + + worden + + + + + + + ; + + + + + + + ſo + + + + + + + ſchaͤdlich + + + + + + + iſt + + + + + + + es + + + + + + + Vorurtheile + + + + + + + zu + + + + + gewiegelt worden; ſo ſchaͤdlich iſt es Vorurtheile zu + + + + + + + + + + pflanzen + + + + + + + , + + + + + + + weil + + + + + + + ſie + + + + + + + ſich + + + + + + + zuletzt + + + + + + + an + + + + + + + denen + + + + + + + ſelbſt + + + + + + + raͤchen + + + + + + + , + + + + + pflanzen, weil ſie ſich zuletzt an denen ſelbſt raͤchen, + + + + + + + + + + die + + + + + + + , + + + + + + + oder + + + + + + + deren + + + + + + + Vorgaͤnger + + + + + + + , + + + + + + + ihre + + + + + + + Urheber + + + + + + + geweſen + + + + + die, oder deren Vorgaͤnger, ihre Urheber geweſen + + + + + + + + + + ſind + + + + + + + . + + + + + + + Daher + + + + + + + kann + + + + + + + ein + + + + + + + Publikum + + + + + + + nur + + + + + + + langſam + + + + + + + zur + + + + + ſind. Daher kann ein Publikum nur langſam zur + + + + + + + + + + Aufklaͤrung + + + + + + + gelangen + + + + + + + . + + + + + + + Durch + + + + + + + eine + + + + + + + Revolution + + + + + + + wird + + + + + Aufklaͤrung gelangen. Durch eine Revolution wird + + + + + + + + + + vielleicht + + + + + + + wohl + + + + + + + ein + + + + + + + Abfall + + + + + + + von + + + + + + + perſoͤnlichem + + + + + + + Despo- + + + + + vielleicht wohl ein Abfall von perſoͤnlichem Despo- + + + + + + + + + + tism + + + + + + + und + + + + + + + gewinnſuͤchtiger + + + + + + + oder + + + + + + + herrſchſüchtiger + + + + + + + Be + + + + + + + - + + + + + tism und gewinnſuͤchtiger oder herrſchſüchtiger Be- + + + + + + + + + + druͤkkung + + + + + + + , + + + + + + + aber + + + + + + + niemals + + + + + + + wahre + + + + + + + Reform + + + + + + + der + + + + + + + Den + + + + + + + - + + + + + druͤkkung, aber niemals wahre Reform der Den- + + + + + + + + + + kungsart + + + + + + + zu + + + + + + + Stande + + + + + + + kommen + + + + + + + ; + + + + + + + ſondern + + + + + + + neue + + + + + + + Vor + + + + + + + - + + + + + kungsart zu Stande kommen; ſondern neue Vor- + + + + + + + + + + urtheile + + + + + + + werden + + + + + + + , + + + + + + + eben + + + + + + + ſowohl + + + + + + + als + + + + + + + die + + + + + + + alten + + + + + + + , + + + + + + + zum + + + + + urtheile werden, eben ſowohl als die alten, zum + + + + + + + + + + Leitbande + + + + + + + des + + + + + + + gedankenloſen + + + + + + + großen + + + + + + + Haufens + + + + + Leitbande des gedankenloſen großen Haufens + + + + + + + + + + dienen + + + + + + + . + + + + + dienen. + + + + gewiegelt worden; ſo ſchaͤdlich iſt es Vorurtheile zu +pflanzen, weil ſie ſich zuletzt an denen ſelbſt raͤchen, +die, oder deren Vorgaͤnger, ihre Urheber geweſen +ſind. Daher kann ein Publikum nur langſam zur +Aufklaͤrung gelangen. Durch eine Revolution wird +vielleicht wohl ein Abfall von perſoͤnlichem Despo- +tism und gewinnſuͤchtiger oder herrſchſüchtiger Be- +druͤkkung, aber niemals wahre Reform der Den- +kungsart zu Stande kommen; ſondern neue Vor- +urtheile werden, eben ſowohl als die alten, zum +Leitbande des gedankenloſen großen Haufens +dienen. + + + + + + + + + + + Zu + + + + + + + dieſer + + + + + + + Aufklaͤrung + + + + + + + aber + + + + + + + wird + + + + + + + nichts + + + + + + + erfordert + + + + + Zu dieſer Aufklaͤrung aber wird nichts erfordert + + + + + + + + + + als + + + + + + + Freiheit + + + + + + + ; + + + + + + + und + + + + + + + zwar + + + + + + + die + + + + + + + unſchaͤdlichſte + + + + + + + unter + + + + + als Freiheit; und zwar die unſchaͤdlichſte unter + + + + + + + + + allem + + + + + + + , + + + + + + + was + + + + + + + nur + + + + + + + Freiheit + + + + + + + heißen + + + + + + + mag + + + + + + + , + + + + + + + naͤmlich + + + + + + + die + + + + + + + : + + + + + allem, was nur Freiheit heißen mag, naͤmlich die: + + + + + + + + + + von + + + + + + + ſeiner + + + + + + + Vernunft + + + + + + + in + + + + + + + allen + + + + + + + Stuͤkken + + + + + + + oͤffentlichen + + + + + von ſeiner Vernunft in allen Stuͤkken oͤffentlichen + + + + + + + + + + Gebrauch + + + + + + + zu + + + + + + + machen + + + + + + + . + + + + + + + Nun + + + + + + + hoͤre + + + + + + + ich + + + + + + + aber + + + + + + + von + + + + + + + al + + + + + + + - + + + + + Gebrauch zu machen. Nun hoͤre ich aber von al- + + + + + + + + + + len + + + + + + + Seiten + + + + + + + rufen + + + + + + + : + + + + + + + raͤſonnirt + + + + + + + nicht + + + + + + + ! + + + + + + + Der + + + + + + + Offi + + + + + + + - + + + + + len Seiten rufen: raͤſonnirt nicht! Der Offi- + + + + + + + + + + zier + + + + + + + ſagt + + + + + + + : + + + + + + + raͤſonnirt + + + + + + + nicht + + + + + + + , + + + + + + + ſondern + + + + + + + exercirt + + + + + + + ! + + + + + + + Der + + + + + zier ſagt: raͤſonnirt nicht, ſondern exercirt! Der + + + + + + + + + + Finanzrath + + + + + + + : + + + + + + + raͤſonnirt + + + + + + + nicht + + + + + + + , + + + + + + + ſondern + + + + + + + bezahlt + + + + + + + ! + + + + + + + Der + + + + + Finanzrath: raͤſonnirt nicht, ſondern bezahlt! Der + + + + + + + + + + Geiſtliche + + + + + + + : + + + + + + + raͤſonnirt + + + + + + + nicht + + + + + + + , + + + + + + + ſondern + + + + + + + glaubt + + + + + + + ! + + + + + + + ( + + + + + + + Nur + + + + + Geiſtliche: raͤſonnirt nicht, ſondern glaubt! (Nur + + + + + + + + + + ein + + + + + + + einziger + + + + + + + Herr + + + + + + + in + + + + + + + der + + + + + + + Welt + + + + + + + ſagt + + + + + + + : + + + + + + + raͤſonnirt + + + + + + + , + + + + + + + ſo + + + + + ein einziger Herr in der Welt ſagt: raͤſonnirt, ſo + + + + + + + + + + viel + + + + + + + ihr + + + + + + + wollt + + + + + + + , + + + + + + + und + + + + + + + woruͤber + + + + + + + ihr + + + + + + + wollt + + + + + + + ; + + + + + + + aber + + + + + + + ge + + + + + + + - + + + + + viel ihr wollt, und woruͤber ihr wollt; aber ge- + + + + + + + + + + horcht + + + + + + + ! + + + + + + + ) + + + + + + + Hier + + + + + + + iſt + + + + + + + uͤberall + + + + + + + Einſchraͤnkung + + + + + + + der + + + + + + + Frei + + + + + + + - + + + + + horcht!) Hier iſt uͤberall Einſchraͤnkung der Frei- + + + + + + + + + + heit + + + + + + + . + + + + + + + Welche + + + + + + + Einſchraͤnkung + + + + + + + aber + + + + + + + iſt + + + + + + + der + + + + + + + Aufklaͤ + + + + + + + - + + + + + heit. Welche Einſchraͤnkung aber iſt der Aufklaͤ- + + + + + + + + + + rung + + + + + + + hinderlich + + + + + + + ? + + + + + + + welche + + + + + + + nicht + + + + + + + , + + + + + + + ſondern + + + + + + + ihr + + + + + + + wohl + + + + + + + gar + + + + + rung hinderlich? welche nicht, ſondern ihr wohl gar + + + + + + + + + + befoͤrderlich + + + + + + + ? + + + + + + + — + + + + + + + Ich + + + + + + + antworte + + + + + + + : + + + + + + + der + + + + + + + oͤffentliche + + + + + befoͤrderlich? — Ich antworte: der oͤffentliche + + + + + + + + + + Gebrauch + + + + + + + ſeiner + + + + + + + Vernunft + + + + + + + muß + + + + + + + jederzeit + + + + + + + frei + + + + + + + ſein + + + + + + + , + + + + + Gebrauch ſeiner Vernunft muß jederzeit frei ſein, + + + + + + + + + + und + + + + + + + der + + + + + + + allein + + + + + + + kann + + + + + + + Aufklaͤrung + + + + + + + unter + + + + + + + Menſchen + + + + + + + zu + + + + + und der allein kann Aufklaͤrung unter Menſchen zu + + + + + Zu dieſer Aufklaͤrung aber wird nichts erfordert +als Freiheit; und zwar die unſchaͤdlichſte unter +allem, was nur Freiheit heißen mag, naͤmlich die: +von ſeiner Vernunft in allen Stuͤkken oͤffentlichen +Gebrauch zu machen. Nun hoͤre ich aber von al- +len Seiten rufen: raͤſonnirt nicht! Der Offi- +zier ſagt: raͤſonnirt nicht, ſondern exercirt! Der +Finanzrath: raͤſonnirt nicht, ſondern bezahlt! Der +Geiſtliche: raͤſonnirt nicht, ſondern glaubt! (Nur +ein einziger Herr in der Welt ſagt: raͤſonnirt, ſo +viel ihr wollt, und woruͤber ihr wollt; aber ge- +horcht!) Hier iſt uͤberall Einſchraͤnkung der Frei- +heit. Welche Einſchraͤnkung aber iſt der Aufklaͤ- +rung hinderlich? welche nicht, ſondern ihr wohl gar +befoͤrderlich? — Ich antworte: der oͤffentliche +Gebrauch ſeiner Vernunft muß jederzeit frei ſein, +und der allein kann Aufklaͤrung unter Menſchen zu + + + + + + + + + + + Stan + + + + + + + - + + + + + Stan- + + + + + Stan- + + + + + + + + + + \ No newline at end of file diff --git a/tests/test_run.py b/tests/test_run.py index 607140e..be928a0 100644 --- a/tests/test_run.py +++ b/tests/test_run.py @@ -1,22 +1,30 @@ from os import environ from pathlib import Path +import pytest import logging from PIL import Image -from eynollah.cli import layout as layout_cli, binarization as binarization_cli +from eynollah.cli import ( + layout as layout_cli, + binarization as binarization_cli, + enhancement as enhancement_cli, + machine_based_reading_order as mbreorder_cli, + ocr as ocr_cli, +) from click.testing import CliRunner from ocrd_modelfactory import page_from_file from ocrd_models.constants import NAMESPACES as NS testdir = Path(__file__).parent.resolve() -EYNOLLAH_MODELS = environ.get('EYNOLLAH_MODELS', str(testdir.joinpath('..', 'models_eynollah').resolve())) -SBBBIN_MODELS = environ.get('SBBBIN_MODELS', str(testdir.joinpath('..', 'default-2021-03-09').resolve())) +MODELS_LAYOUT = environ.get('MODELS_LAYOUT', str(testdir.joinpath('..', 'models_layout_v0_5_0').resolve())) +MODELS_OCR = environ.get('MODELS_OCR', str(testdir.joinpath('..', 'models_ocr_v0_5_0').resolve())) +MODELS_BIN = environ.get('MODELS_BIN', str(testdir.joinpath('..', 'default-2021-03-09').resolve())) def test_run_eynollah_layout_filename(tmp_path, subtests, pytestconfig, caplog): infile = testdir.joinpath('resources/kant_aufklaerung_1784_0020.tif') outfile = tmp_path / 'kant_aufklaerung_1784_0020.xml' args = [ - '-m', EYNOLLAH_MODELS, + '-m', MODELS_LAYOUT, '-i', str(infile), '-o', str(outfile.parent), # subtests write to same location @@ -44,8 +52,7 @@ def test_run_eynollah_layout_filename(tmp_path, subtests, pytestconfig, caplog): options=options): with caplog.filtering(only_eynollah): result = runner.invoke(layout_cli, args + options, catch_exceptions=False) - print(result) - assert result.exit_code == 0 + assert result.exit_code == 0, result.stdout logmsgs = [logrec.message for logrec in caplog.records] assert str(infile) in logmsgs assert outfile.exists() @@ -61,7 +68,7 @@ def test_run_eynollah_layout_directory(tmp_path, pytestconfig, caplog): indir = testdir.joinpath('resources') outdir = tmp_path args = [ - '-m', EYNOLLAH_MODELS, + '-m', MODELS_LAYOUT, '-di', str(indir), '-o', str(outdir), ] @@ -72,9 +79,8 @@ def test_run_eynollah_layout_directory(tmp_path, pytestconfig, caplog): return logrec.name == 'eynollah' runner = CliRunner() with caplog.filtering(only_eynollah): - result = runner.invoke(layout_cli, args) - print(result) - assert result.exit_code == 0 + result = runner.invoke(layout_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout logmsgs = [logrec.message for logrec in caplog.records] assert len([logmsg for logmsg in logmsgs if logmsg.startswith('Job done in')]) == 2 assert any(logmsg for logmsg in logmsgs if logmsg.startswith('All jobs done in')) @@ -84,10 +90,12 @@ def test_run_eynollah_binarization_filename(tmp_path, subtests, pytestconfig, ca infile = testdir.joinpath('resources/kant_aufklaerung_1784_0020.tif') outfile = tmp_path.joinpath('kant_aufklaerung_1784_0020.png') args = [ - '-m', SBBBIN_MODELS, - str(infile), - str(outfile), + '-m', MODELS_BIN, + '-i', str(infile), + '-o', str(outfile), ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) caplog.set_level(logging.INFO) def only_eynollah(logrec): return logrec.name == 'SbbBinarizer' @@ -99,9 +107,8 @@ def test_run_eynollah_binarization_filename(tmp_path, subtests, pytestconfig, ca with subtests.test(#msg="test CLI", options=options): with caplog.filtering(only_eynollah): - result = runner.invoke(binarization_cli, args + options) - print(result) - assert result.exit_code == 0 + result = runner.invoke(binarization_cli, args + options, catch_exceptions=False) + assert result.exit_code == 0, result.stdout logmsgs = [logrec.message for logrec in caplog.records] assert any(True for logmsg in logmsgs if logmsg.startswith('Predicting')) assert outfile.exists() @@ -115,18 +122,193 @@ def test_run_eynollah_binarization_directory(tmp_path, subtests, pytestconfig, c indir = testdir.joinpath('resources') outdir = tmp_path args = [ - '-m', SBBBIN_MODELS, + '-m', MODELS_BIN, '-di', str(indir), - '-do', str(outdir), + '-o', str(outdir), ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) caplog.set_level(logging.INFO) def only_eynollah(logrec): return logrec.name == 'SbbBinarizer' runner = CliRunner() with caplog.filtering(only_eynollah): - result = runner.invoke(binarization_cli, args) - print(result) - assert result.exit_code == 0 + result = runner.invoke(binarization_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout logmsgs = [logrec.message for logrec in caplog.records] assert len([logmsg for logmsg in logmsgs if logmsg.startswith('Predicting')]) == 2 assert len(list(outdir.iterdir())) == 2 + +def test_run_eynollah_enhancement_filename(tmp_path, subtests, pytestconfig, caplog): + infile = testdir.joinpath('resources/kant_aufklaerung_1784_0020.tif') + outfile = tmp_path.joinpath('kant_aufklaerung_1784_0020.png') + args = [ + '-m', MODELS_LAYOUT, + '-i', str(infile), + '-o', str(outfile.parent), + # subtests write to same location + '--overwrite', + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.INFO) + def only_eynollah(logrec): + return logrec.name == 'enhancement' + runner = CliRunner() + for options in [ + [], # defaults + ["-sos"], + ]: + with subtests.test(#msg="test CLI", + options=options): + with caplog.filtering(only_eynollah): + result = runner.invoke(enhancement_cli, args + options, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + assert any(True for logmsg in logmsgs if logmsg.startswith('Image was enhanced')), logmsgs + assert outfile.exists() + with Image.open(infile) as original_img: + original_size = original_img.size + with Image.open(outfile) as enhanced_img: + enhanced_size = enhanced_img.size + assert (original_size == enhanced_size) == ("-sos" in options) + +def test_run_eynollah_enhancement_directory(tmp_path, subtests, pytestconfig, caplog): + indir = testdir.joinpath('resources') + outdir = tmp_path + args = [ + '-m', MODELS_LAYOUT, + '-di', str(indir), + '-o', str(outdir), + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.INFO) + def only_eynollah(logrec): + return logrec.name == 'enhancement' + runner = CliRunner() + with caplog.filtering(only_eynollah): + result = runner.invoke(enhancement_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + assert len([logmsg for logmsg in logmsgs if logmsg.startswith('Image was enhanced')]) == 2 + assert len(list(outdir.iterdir())) == 2 + +def test_run_eynollah_mbreorder_filename(tmp_path, subtests, pytestconfig, caplog): + infile = testdir.joinpath('resources/kant_aufklaerung_1784_0020.xml') + outfile = tmp_path.joinpath('kant_aufklaerung_1784_0020.xml') + args = [ + '-m', MODELS_LAYOUT, + '-i', str(infile), + '-o', str(outfile.parent), + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.INFO) + def only_eynollah(logrec): + return logrec.name == 'mbreorder' + runner = CliRunner() + with caplog.filtering(only_eynollah): + result = runner.invoke(mbreorder_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + # FIXME: mbreorder has no logging! + #assert any(True for logmsg in logmsgs if logmsg.startswith('???')), logmsgs + assert outfile.exists() + #in_tree = page_from_file(str(infile)).etree + #in_order = in_tree.xpath("//page:OrderedGroup//@regionRef", namespaces=NS) + out_tree = page_from_file(str(outfile)).etree + out_order = out_tree.xpath("//page:OrderedGroup//@regionRef", namespaces=NS) + #assert len(out_order) >= 2, "result is inaccurate" + #assert in_order != out_order + assert out_order == ['r_1_1', 'r_2_1', 'r_2_2', 'r_2_3'] + +def test_run_eynollah_mbreorder_directory(tmp_path, subtests, pytestconfig, caplog): + indir = testdir.joinpath('resources') + outdir = tmp_path + args = [ + '-m', MODELS_LAYOUT, + '-di', str(indir), + '-o', str(outdir), + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.INFO) + def only_eynollah(logrec): + return logrec.name == 'mbreorder' + runner = CliRunner() + with caplog.filtering(only_eynollah): + result = runner.invoke(mbreorder_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + # FIXME: mbreorder has no logging! + #assert len([logmsg for logmsg in logmsgs if logmsg.startswith('???')]) == 2 + assert len(list(outdir.iterdir())) == 2 + +def test_run_eynollah_ocr_filename(tmp_path, subtests, pytestconfig, caplog): + infile = testdir.joinpath('resources/kant_aufklaerung_1784_0020.tif') + outfile = tmp_path.joinpath('kant_aufklaerung_1784_0020.xml') + outrenderfile = tmp_path.joinpath('render').joinpath('kant_aufklaerung_1784_0020.png') + outrenderfile.parent.mkdir() + args = [ + '-m', MODELS_OCR, + '-i', str(infile), + '-dx', str(infile.parent), + '-o', str(outfile.parent), + # subtests write to same location + '--overwrite', + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.DEBUG) + def only_eynollah(logrec): + return logrec.name == 'eynollah' + runner = CliRunner() + for options in [ + # kba Fri Sep 26 12:53:49 CEST 2025 + # Disabled until NHWC/NCHW error in https://github.com/qurator-spk/eynollah/actions/runs/18019655200/job/51273541895 debugged + # [], # defaults + # ["-doit", str(outrenderfile.parent)], + ["-trocr"], + ]: + with subtests.test(#msg="test CLI", + options=options): + with caplog.filtering(only_eynollah): + result = runner.invoke(ocr_cli, args + options, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + # FIXME: ocr has no logging! + #assert any(True for logmsg in logmsgs if logmsg.startswith('???')), logmsgs + assert outfile.exists() + if "-doit" in options: + assert outrenderfile.exists() + #in_tree = page_from_file(str(infile)).etree + #in_order = in_tree.xpath("//page:OrderedGroup//@regionRef", namespaces=NS) + out_tree = page_from_file(str(outfile)).etree + out_texts = out_tree.xpath("//page:TextLine/page:TextEquiv[last()]/page:Unicode/text()", namespaces=NS) + assert len(out_texts) >= 2, ("result is inaccurate", out_texts) + assert sum(map(len, out_texts)) > 100, ("result is inaccurate", out_texts) + +@pytest.mark.skip("Disabled until NHWC/NCHW error in https://github.com/qurator-spk/eynollah/actions/runs/18019655200/job/51273541895 debugged") +def test_run_eynollah_ocr_directory(tmp_path, subtests, pytestconfig, caplog): + indir = testdir.joinpath('resources') + outdir = tmp_path + args = [ + '-m', MODELS_OCR, + '-di', str(indir), + '-dx', str(indir), + '-o', str(outdir), + ] + if pytestconfig.getoption('verbose') > 0: + args.extend(['-l', 'DEBUG']) + caplog.set_level(logging.INFO) + def only_eynollah(logrec): + return logrec.name == 'eynollah' + runner = CliRunner() + with caplog.filtering(only_eynollah): + result = runner.invoke(ocr_cli, args, catch_exceptions=False) + assert result.exit_code == 0, result.stdout + logmsgs = [logrec.message for logrec in caplog.records] + # FIXME: ocr has no logging! + #assert any(True for logmsg in logmsgs if logmsg.startswith('???')), logmsgs + assert len(list(outdir.iterdir())) == 2