Commits · 2b2a90a2a01100214c452baa6224f1d29387b5f3 · Fabian Mersch / SimpleHTR

Browse files

Mar 29, 2023
- Fixed Bugs and Added Train Loss Dump Feature · 2b2a90a2
  Alexander Alexeev authored 2 years ago
  
  2b2a90a2
Mar 01, 2022
- in validation mode use charset from trained model and not from dataset, see issue #127 · d807bc44
  Harald Scheidl authored 3 years ago
  
  d807bc44
May 27, 2021
- readme, docstrings, and other small changes · ab6220d6
  Harald Scheidl authored 4 years ago
  
  ab6220d6
May 25, 2021
- ready for first line-level training runs · 2b0b5484
  Harald Scheidl authored 4 years ago
  
  2b0b5484
May 24, 2021
- reworked preprocessor · 3507e59d
  Harald Scheidl authored 4 years ago
  
  3507e59d
- started multi word training · 6c76bca2
  Harald Scheidl authored 4 years ago
  
  6c76bca2
- allow dynamic width of image · 7edd1f86
  Harald Scheidl authored 4 years ago
  
  7edd1f86
- changed code to python coding style · 8fd81536
  Harald Scheidl authored 4 years ago
  
  8fd81536
Feb 03, 2021
- single option for decoders · a7e85ba7
  Harald Scheidl authored 4 years ago
  
  a7e85ba7
Feb 02, 2021
- log both word acc and cer for valset · 00d7232d
  Harald Scheidl authored 4 years ago
  
  00d7232d
- data aug: random background, write summary of CERs while training · cdfb5188
  Harald Scheidl authored 4 years ago
  
  cdfb5188
Jan 22, 2021
- reworked readme, optimized imports · 10366c39
  Harald Scheidl authored 4 years ago
  
  10366c39
Jan 20, 2021
- reworked code: faster dataloader, more data augmentations, and some more changes... · f3945bd7
  Harald Scheidl authored 4 years ago
  
  f3945bd7
Dec 10, 2020
- rm unused import · 87817b79
  Harald Scheidl authored 4 years ago
  
  87817b79
- reformat code, add note about TF2 in readme · 45162d8a
  Harald Scheidl authored 4 years ago
  
  45162d8a
Dec 08, 2020
- Fixed inferBatch ValueError. · 92b4f31c
  Nishant authored 4 years ago
  
  92b4f31c
May 25, 2019
- added dump option for NN output · e8249f7c
  Harald Scheidl authored 6 years ago
  
  e8249f7c
Mar 15, 2019
- unify code style · f2f606da
  Harald Scheidl authored 6 years ago
  
  f2f606da
Jan 02, 2019
- dynamic batch size · f869e085
  Harald Scheidl authored 6 years ago
  
  f869e085
- calc probability for recognized text · c1ef4818
  Harald Scheidl authored 6 years ago
  
  c1ef4818
Nov 27, 2018
- py2 compatibility (print, div) via __future__ · 50a4c6e7
  Harald Scheidl authored 6 years ago
  
  50a4c6e7
- Addressed cv2.imread terminations and python 2 cross-compatibility errors. · 88d7eca2
  Fake Name authored 6 years ago
  
  88d7eca2
Aug 30, 2018
- integrate word beam search decoding · 97192a92
  Harald Scheidl authored 6 years ago
  
  97192a92
Aug 17, 2018
- decaying learning rate, changed early stopping parameter · 998cc601
  Harald Scheidl authored 6 years ago
  
  998cc601
- show CER and accuracy · 0f0947b8
  Harald Scheidl authored 6 years ago
  
  0f0947b8
Aug 16, 2018
- using character error rate instead of word accuracy · b9fc4e3a
  Harald Scheidl authored 6 years ago
  
  b9fc4e3a
Jul 24, 2018
- validation: removed trainBatch · a451e5f0
  Harald Scheidl authored 6 years ago
  
  a451e5f0
- validation mode, optional beam search · 617e9e28
  Harald Scheidl authored 6 years ago
  
  617e9e28
Jun 19, 2018
- use random subset of training set for faster epochs · 67859b2c
  Harald Scheidl authored 7 years ago
  
  67859b2c
Jun 18, 2018
- data augmentation while training · df7cb0a8
  Harald Scheidl authored 7 years ago
  
  df7cb0a8
Jun 15, 2018
- training with early stopping algo · c9db6ea8
  Harald Scheidl authored 7 years ago
  
  c9db6ea8
Jun 13, 2018
- check that there is a saved model in inference mode · da316a35
  Harald Scheidl authored 7 years ago
  
  da316a35
- changed cli · b3962fd1
  Harald Scheidl authored 7 years ago
  
  b3962fd1
- init · 39c4684b
  Harald Scheidl authored 7 years ago
  
  39c4684b