qurator-spk
/
eynollah
mirror of https://github.com/qurator-spk/eynollah.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
410 Commits
14 Branches
16 Tags
16 MiB
Commit Graph

152 Commits (48f2ce62034bbabefe909820ad8b9cee2ebda47f)

Author SHA1 Message Date
vahidrezanezhad 1621532092
Merge branch 'main' into eynollah_light
vahid 4c217018cc textline light version -tll can not work without enabling -light option
vahid d68f240b59 loading TensorFlow SavedModel format is now present
vahid 380f59ad67 let hybrid textline light model be loaded
vahidrezanezhad 29e6ad076f
renaming textline light model
Robert Sachunsky 529f2c0e19 set_memory_growth to all GPU devices alike
Robert Sachunsky 8fe3567123
set_memory_growth to all GPU devices alike
vahid d3735b12f4 pushing commits 2d9ccac and 7345f6b into eynollah_light
cneud 817e5a6af9 update docstring
cneud a9728bb899 Update eynollah.py 
predict quietly please
Robert Sachunsky fb6d97091b OCR-D wrapper: expose tables param
Robert Sachunsky 1ac0a7e06f try loading as TF SavedModel instead of HDF5
Robert Sachunsky 73057d57d1 silentium!
cneud 3d54719c87 fix import
cneud 58ca226f2d apply some fixes from main
Konstantin Baierer 6c8f60993f Merge branch 'main' into 420-namespace-package
Konstantin Baierer ea792d1e4a 📦 v0.2.0
Konstantin Baierer 8e894e5b7b Merge remote-tracking branch 'bertsky/savedmodel-format'
Konstantin Baierer 7cd07dd550 use PEP420 style qurator namespace
Konstantin Baierer 71d0ec8dfe 📦 v0.1.0
Robert Sachunsky 875e4fe32b log number of detected regions
Robert Sachunsky 318ea6acca OCR-D wrapper: expose tables param
Robert Sachunsky 23f0c0b40a ocrd-tool: replace by persistent model URL
Robert Sachunsky 5c26bdf402 ocrd-tool: add model archive to resmgr resources
Robert Sachunsky 7345f6bf67 remove TF1 session and GC controls, avoid repeating load_model
Robert Sachunsky a56988a35a contours: numpy now needs dtype=object
Robert Sachunsky 2d9ccac354 contours: simplify
Robert Sachunsky ab4bb7cd7b silentium!
Robert Sachunsky 79e897d3b2 try loading as TF SavedModel instead of HDF5
vahid 38bf0d8740 solving issue by loading model by directory as input
vahid 583cdcee2c new (hybrid cnn+transformer) textline model which can accelerate to extract contour textlines faster
vahid 8d5079c909 issue is resolved on main branch
vahid 402c5339ac issue is resolved
vahid 01bfc3914d extracting page as an option
vahidrezanezhad ae7c424889
Update eynollah.py
vahid cd9920eea7 extracting page
vahid 735abc43f3 option to ignore page extraction
cneud 934bbd5892 cleanup
cneud ecf117ca95 adapt to tf1.compat session mode in tf2
Clemens Neudecker 568391ec4a
require model command line option (fix ) ()
vahid 3bbbeecfec all options are enabled for light version
Gerber, Mike f27ac155ae 🧹 Downgrade "Patch size" log message to debug 
Fixes gh-55.
vahid adf10942fa issue resolved
vahid 2eacb9a8ec renaming the models
vahid c606391c31 flow from directory
vahid cf5ef8f5ae light version as option
vahidrezanezhad b8a532180a light version integration
vahidrezanezhad 10f1acef29
Merge pull request from mikegerber/fix/enhanced-message 
Fix/enhanced message
vahidrezanezhad c30d4d5c30
Merge pull request from mikegerber/feat/better-time-msgs 
💄 Improve timing messages (Fixes )
Gerber, Mike 11d9b00510 🧹 Don't produce spurious TextEquiv elements. 
eynollah produces spurious - and empy - pcGts TextEquiv elements. This
is a. unnecessary, b. wrong and c. produces a lot of warning messages
in subsequent OCR processing steps because the OCR processor warns
about already existing text.

Fix this by not generating any TextEquiv elements.

Fixes gh-37.