Robert Sachunsky
4f7c5675fc
ocr: ensure forked logging handlers also work under pytest
2026-07-18 18:02:40 +02:00
Robert Sachunsky
79a9bb0128
cnn-rnn-ocr: increase default batch size and VRAM limit
2026-07-18 00:59:57 +02:00
Robert Sachunsky
be8b161607
cnn-rnn-ocr: batch flipped line candidates together with rest
2026-07-18 00:37:19 +02:00
Robert Sachunsky
7e776612a4
cnn-rnn-ocr: if dir_in_bin==dir_in, then split PNG and rest…
...
(supports common case that binarized images have same stem,
but different file name extension)
2026-07-18 00:35:50 +02:00
Robert Sachunsky
6840b67961
ocr: run dir_in mode in parallel (like layout), too…
...
- add CLI options `--num-jobs` and `--halt-fail`
- separate `Eynollah_ocr.run_single()` to be scheduled
- use ProcessPoolExecutor w/ forking and QueueListener
- also skip if input XML file is missing
- log processing times per job and overall
2026-07-18 00:24:08 +02:00
Robert Sachunsky
1c8ac38d31
OCR w/o overwrite: skip one file, not the entire run
2026-07-17 19:19:20 +02:00
Robert Sachunsky
aace571368
cnn-rnn-ocr: fix
2026-07-17 19:15:54 +02:00
Robert Sachunsky
5d129dc8c1
cnn-rnn-ocr: fix typo causing rare failures
2026-07-17 19:02:42 +02:00
Robert Sachunsky
4b9fa543ae
processor: fix typo (empty model_overrides)
2026-07-17 16:23:03 +02:00
Robert Sachunsky
5939845d1d
utils.contour.make_valid: be more robust (avoiding MultiPolygon)
2026-07-17 15:43:47 +02:00
Robert Sachunsky
c1b276fea1
processor: pass on more Eynollah parameters…
...
- `device` selection
- `model_overrides` (as dict; including relative path resolution), e.g.
```JSON
{
"binarization": {
"":
"models_inference_layout_v0_9_1/models_eynollah/eynollah-binarization_20210425.onnx"
}
}
```
- `skip_layout_and_reading_order`
- `num_col_upper`
- `num_col_lower`
- `binarize` (for `input_binary`, which is a misnomer)
2026-07-17 15:37:39 +02:00
Robert Sachunsky
f579d12866
processor: resolve models path as processor resource
2026-07-17 12:56:19 +02:00
Robert Sachunsky
0a9b3097f1
processor: avoid writing XML twice (once by writer, once by OCR-D)
2026-07-17 12:53:13 +02:00
kba
d2755d1e93
update references to new v0_9_1 model release
2026-07-15 20:05:36 +02:00
kba
1440d454cc
models: bump version
2026-07-15 17:18:33 +02:00
kba
32c5e9ae76
add missing models, remove microsoft
2026-07-15 17:03:04 +02:00
kba
03915b24a5
models: symlinks for the model packages for zenodo upload
2026-07-15 16:02:49 +02:00
kba
909ccfd38b
Merge branch 'fix-0.8-modelzoo-and-predictor' of https://github.com/bertsky/eynollah into fix-0.8-modelzoo-and-predictor
2026-07-15 15:37:11 +02:00
Robert Sachunsky
9804d736ac
CI: avoid CUDA dependencies here
2026-07-15 13:59:43 +02:00
Robert Sachunsky
4d97e3bf7f
training: fix typos found by ruff
2026-07-15 12:46:33 +02:00
Robert Sachunsky
0ab6e19f33
configure ruff search path
2026-07-15 12:46:13 +02:00
Robert Sachunsky
b89e1b4296
ocrd-tool.json: differentiate inference-all and inference-layout
2026-07-15 04:24:33 +02:00
Robert Sachunsky
25865372d0
dependencies for [OCR]: add TF, avoid newer Torch (pulling CUDA 13)
2026-07-15 04:11:29 +02:00
Robert Sachunsky
efe4aa8b0c
mbreorder: fix init (wrong way to load model)
2026-07-15 04:10:47 +02:00
Robert Sachunsky
839d7b9a7e
ModelZoo: avoid Azure EP (displacing CPU)
2026-07-15 04:09:33 +02:00
kba
c5713e010e
deps: OCR requires explicit dep on tensorflow/keras now
2026-07-14 21:04:06 +02:00
kba
b4165114ea
update zenodo links to v0_9_0 models
2026-07-14 21:00:09 +02:00
kba
c43e219858
model packaging/uploading
2026-07-14 20:51:46 +02:00
Robert Sachunsky
83fca95914
Merge pull request #9 from qurator-spk/integrating_trocr_and_torch_ensembling_and_updating_characters_list-refactor
...
Integrating trocr and torch ensembling and updating characters list refactor
2026-07-14 16:23:12 +02:00
Robert Sachunsky
1591d2091c
update docs
2026-07-14 16:22:47 +02:00
Robert Sachunsky
5dc9a3456c
Dockerfile: do not install OCR (as Torch and ONNX clash over CUDA)
2026-07-14 15:58:00 +02:00
Robert Sachunsky
12be983487
cnn-rnn-ocr inference: switch back to beam search, only run in TF…
...
- training.models.CTCDecoder: prefer beam search over greedy
(because it is more accurate)
- training reload makefile: skip ONNX and TF-Serving conversion
for cnn-rnn-ocr models (because these would not work)
- training reload makefile: default to onnx and tf conversions
for all models (tf for training and onnx for inference)
instead of tf-serving export
2026-07-12 03:49:41 +02:00
Robert Sachunsky
c680dae2d1
ModelZoo for ONNX backend: allow setting execution providers via env
2026-07-10 17:22:50 +02:00
kba
39e054e718
train: remove tf-specifics and weird __getitem__ def from transformer-ocr setup
2026-07-09 19:23:21 +02:00
kba
affddd6c85
train: make preprocess_imgs_ocr work for transformer-ocr
2026-07-09 19:12:09 +02:00
kba
b2f3a8f2d8
Merge branch 'fix-0.8-modelzoo-and-predictor-kba0709' into integrating_trocr_and_torch_ensembling_and_updating_characters_list-refactor
...
# Conflicts:
# train/requirements.txt
2026-07-09 17:32:23 +02:00
kba
b1f2f43051
upgrade tf2onnx fork dep, remove spurious tf-data dep
2026-07-09 17:30:48 +02:00
kba
492fcbacb7
switch to fork of tf2onnx
2026-07-09 15:17:11 +02:00
kba
b9ba43b444
training: require tf2onnx and pin ml_dtypes >= 0.5
2026-07-09 14:37:51 +02:00
Robert Sachunsky
c9c14ed83d
Docker: update to ONNX base image (in Makefile, too)
2026-07-08 14:04:19 +02:00
Robert Sachunsky
98e28ca9f4
Docker: fixup cuDNN installation after OCR extra
2026-07-08 03:05:07 +02:00
Robert Sachunsky
5354583913
update readme
2026-07-08 02:56:11 +02:00
Robert Sachunsky
8cc8c28471
ModelZoo for ONNX backend: configure TRT cache path from XDG env
2026-07-08 02:55:25 +02:00
Robert Sachunsky
32568a590f
Docker: update to ONNX base image
2026-07-08 02:54:34 +02:00
Robert Sachunsky
171a8a3161
move TF+Keras dependencies to training extra, replace by ONNX+TRT
2026-07-07 22:56:36 +02:00
Robert Sachunsky
1567df1379
remove numba dependency (previously used to free CUDA memory)
2026-07-07 22:09:46 +02:00
Robert Sachunsky
16943f70b4
models (ViT backbone) iterate extract_patches over batch dim…
...
`Patches.call`: use `tf.map_fn` instead of running
entire batch through `tf.image.extract_patches`
(faster, less VRAM, allows ONNX conversion to work)
2026-07-02 20:58:22 +02:00
Robert Sachunsky
1b27c7390f
training.convert/ONNX: run strict shape inference and check model
2026-07-02 20:56:55 +02:00
Robert Sachunsky
5e531ab006
get_textlines_of_textregion_sorted: fix be61875
2026-07-01 18:27:46 +02:00
Robert Sachunsky
01e69a0e22
predictor for OCR models: work around ONNX bug …
...
(ONNX converted models already return `np.dtype=object`
arrays of `np.str_` instead of `np.bytes_`; so undo this)
2026-06-26 02:42:51 +02:00
