Commits · d807bc44ea7d54dc7e7d9d407b61ed72a770f64a · Fabian Mersch / SimpleHTR

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
Feb 02, 2021
- all model files are now in the uploaded zip · 66a9f1b9
  Harald Scheidl authored 4 years ago
  
  66a9f1b9
- data aug: random background, write summary of CERs while training · cdfb5188
  Harald Scheidl authored 4 years ago
  
  cdfb5188
Jan 22, 2021
- update accuracy on full val set · 7bd8ca85
  Harald Scheidl authored 4 years ago
  
  7bd8ca85
Jan 21, 2021
- restructured repo, reworked readme, mv model to dropbox · b2df8508
  Harald Scheidl authored 4 years ago
  
  b2df8508
Jan 11, 2019
- updated README: improved accuracy · 2e55e119
  Harald Scheidl authored 6 years ago
  
  2e55e119
Jan 09, 2019
- Added batch normalization and updated model.zip · 7e0da952
  Chazzz authored 6 years ago
  
  7e0da952
Aug 30, 2018
- integrate word beam search decoding · 97192a92
  Harald Scheidl authored 6 years ago
  
  97192a92
Aug 20, 2018
- new pre-trained model · 376c8f4b
  Harald Scheidl authored 6 years ago
  
  376c8f4b
Jun 20, 2018
- decreased CNN kernel size to increase accuracy · f1754076
  Harald Scheidl authored 7 years ago
  
  f1754076
Jun 19, 2018
- use random subset of training set for faster epochs · 67859b2c
  Harald Scheidl authored 7 years ago
  
  67859b2c
Jun 13, 2018
- init · 39c4684b
  Harald Scheidl authored 7 years ago
  
  39c4684b