Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 12, 2024

• adapted the image preprocessing functionds for my dataset

  anuj018

  committedFeb 12, 2024
  1cc10c5

Commits on Feb 11, 2024

• Model is now trainable for line data

  anuj018

  committedFeb 11, 2024
  ae7ca15

Commits on Feb 10, 2024

• converted images to the right shape that the model expects and changed the way sparsetensor output from to_spare() was processed to compute the lengths of the ground trtuh texts needed for CTC loss

  anuj018

  committedFeb 10, 2024
  8ad5184

Commits on Feb 8, 2024

• changed the dataloader_iam create_lmbd,py and mdoel.py to adapt to a different dataset

  anuj018

  committedFeb 8, 2024
  3b4b63c

Commits on Feb 7, 2024

• .

  anuj018

  committedFeb 7, 2024
  9ba4bef

• modified model.py

  anuj018

  committedFeb 7, 2024
  9042928

Commits on Feb 6, 2024

• modified model.py to allow graph mode execution of the model

  anuj018

  committedFeb 6, 2024
  7884e84

Commits on Feb 5, 2024

• porting it to tensorflow version 2.X

  anuj018

  committedFeb 5, 2024
  caf2b83

Commits on Sep 7, 2023

• Merge pull request githubharald#164 from tosemml/tosemml-patch-1

  githubharald

  committedSep 7, 2023
  6342401

Commits on Sep 6, 2023

• Fix error

  tosemml

  committedSep 6, 2023
  64676fa

Commits on Aug 22, 2023

• use join

  tosemml

  committedAug 22, 2023
  92177d3

Commits on Apr 27, 2023

• added link to demo

  

  Harald Scheidl

  committedApr 27, 2023
  77d2a05

Commits on Apr 3, 2023

• Merge pull request githubharald#157 from alexeevrevan/master

  githubharald

  committedApr 3, 2023
  ec1ce16

Commits on Mar 29, 2023

• Fixed Bugs and Added Train Loss Dump Feature

  alexeevrevan

  committedMar 29, 2023
  2b2a90a

Commits on Mar 28, 2023

• added link to repo that uses this model

  

  Harald Scheidl

  committedMar 28, 2023
  2c14ec8

Commits on Oct 13, 2022

• ignore empty lines and comment lines in IAM dataset file

  

  Harald Scheidl

  committedOct 13, 2022
  8577a9e

Commits on Mar 1, 2022

• in validation mode use charset from trained model and not from dataset, see issue githubharald#127

  

  Harald Scheidl

  committedMar 1, 2022
  d807bc4

Commits on Jan 10, 2022

• png images: white instead of transparent bg to handle dark mode

  

  Harald Scheidl

  committedJan 10, 2022
  7f26b32

Commits on May 27, 2021

• readme, docstrings, and other small changes

  

  Harald Scheidl

  committedMay 27, 2021
  ab6220d

Commits on May 25, 2021

• ready for first line-level training runs

  

  Harald Scheidl

  committedMay 25, 2021
  2b0b548

Commits on May 24, 2021

• use correct batch size

  

  Harald Scheidl

  committedMay 24, 2021
  dca7cab

• reworked preprocessor

  

  Harald Scheidl

  committedMay 24, 2021
  3507e59

• started multi word training

  

  Harald Scheidl

  committedMay 24, 2021
  6c76bca

• allow dynamic width of image

  

  Harald Scheidl

  committedMay 24, 2021
  7edd1f8

• changed code to python coding style

  

  Harald Scheidl

  committedMay 24, 2021
  8fd8153

Commits on Feb 18, 2021

• rm old faq section

  

  Harald Scheidl

  committedFeb 18, 2021
  8ba5e99

Commits on Feb 17, 2021

• word beam search: switched from TF custom op to Python package, so it can also be used in Windows now

  

  Harald Scheidl

  committedFeb 17, 2021
  a91b463

Commits on Feb 16, 2021

• cleanup: remove analyze code, and only keep code really relevant for HTR

  

  Harald Scheidl

  committedFeb 16, 2021
  330c3db

Commits on Feb 3, 2021

• single option for decoders

  

  Harald Scheidl

  committedFeb 3, 2021
  a7e85ba

Commits on Feb 2, 2021

• log both word acc and cer for valset

  

  Harald Scheidl

  committedFeb 2, 2021
  00d7232

• changed data aug

  

  Harald Scheidl

  committedFeb 2, 2021
  8c96a80

• data aug: dont move img around if it is already larger than target img

  

  Harald Scheidl

  committedFeb 2, 2021
  5e191e3

• all model files are now in the uploaded zip

  

  Harald Scheidl

  committedFeb 2, 2021
  66a9f1b

• data aug: random background, write summary of CERs while training

  

  Harald Scheidl

  committedFeb 2, 2021
  cdfb518

Commits on Feb 1, 2021

• test: data aug=true

  

  Harald Scheidl

  committedFeb 1, 2021
  2f0499f

Pagination

PreviousNext