Commits

Breadcrumbs

History for

onmaster

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 3, 2019

• Added copyright headers to the source files.

  Show description for b59749c
  gwenniger

  committed
  b59749c

  Copy full SHA for b59749c

Commits on Jul 17, 2019

• Improved the code so that when saving pre-processed examples to individual files, the program looks for previously

  Show description for b57246d
  gwenniger

  committed
  b57246d

  Copy full SHA for b57246d

Commits on Jul 5, 2019

• Fixed the dedicated blank symbol used in string_to_index_mapping_table.

  Show description for 9044fa9
  gwenniger

  committed
  9044fa9

  Copy full SHA for 9044fa9

Commits on Jul 4, 2019

• Added code for improving the images based on the example from

  Show description for 263d8fb
  gwenniger

  committed
  263d8fb

  Copy full SHA for 263d8fb

• Added the functionality of saving to and loading from pre-processed

  Show description for 2b05334
  gwenniger

  committed
  2b05334

  Copy full SHA for 2b05334

Commits on Jun 25, 2019

• Improved the README file.

  Show description for b89714b
  gwenniger

  committed
  b89714b

  Copy full SHA for b89714b

Commits on Jun 24, 2019

• Further extended and improved the Rimes data preparation code.

  Show description for 2a753f4
  gwenniger

  committed
  2a753f4

  Copy full SHA for 2a753f4

Commits on Jun 21, 2019

• Added the functionality to build a dictionary of RimesLineInformation

  Show description for 41e6783
  gwenniger

  committed
  41e6783

  Copy full SHA for 41e6783

Commits on Jun 20, 2019

• Implemented a line_strip annotation creator, which creates a file with pairs of line strip file path

  Show description for 46ab4d9
  gwenniger

  committed
  46ab4d9

  Copy full SHA for 46ab4d9

Commits on Jun 19, 2019

• Added a new clase called line_strip_extractor which takes care of

  Show description for 7df6bff
  gwenniger

  committed
  7df6bff

  Copy full SHA for 7df6bff

• Started the implementation of preprocessing code for the Rimes

  Show description for af2861a
  gwenniger

  committed
  af2861a

  Copy full SHA for af2861a

Commits on Feb 5, 2019

• Completed the maximum memory GPU memory consumption bookkeeping using

  Show description for 39b330e
  

  Gideon Maillette de Buy Wenniger

  committed
  39b330e

  Copy full SHA for 39b330e

Commits on Dec 9, 2018

• Restored the minimal vertical padding, which should work better for lines,

  Show description for f558b90
  gwenniger

  committed
  f558b90

  Copy full SHA for f558b90

Commits on Dec 8, 2018

• Added an optimization of the horizontal padding, which can be reduced

  Show description for e6aefdd
  gwenniger

  committed
  e6aefdd

  Copy full SHA for e6aefdd

Commits on Nov 16, 2018

• Adapted iam_examples_dictionary so that "error: examples:

  Show description for 3ef800e
  gwenniger

  committed
  3ef800e

  Copy full SHA for 3ef800e

Commits on Nov 12, 2018

• Fixed the code in modules/train_multi_dimensional_rnn_ctc.py

  Show description for 04fa22e
  gwenniger

  committed
  04fa22e

  Copy full SHA for 04fa22e

Commits on Nov 1, 2018

• Added code for saving and loading the IAM dataset in its python

  Show description for 39664d3
  gwenniger

  committed
  39664d3

  Copy full SHA for 39664d3

Commits on Oct 22, 2018

• Implemented a way to use 2 by 2 pixel blocks extracted from IAM

  Show description for cd47ea2
  gwenniger

  committed
  cd47ea2

  Copy full SHA for cd47ea2

Commits on Oct 12, 2018

• Fixed one bug and one conceptual error in the IAM dataset creation:

  Show description for 78277f2
  gwenniger

  committed
  78277f2

  Copy full SHA for 78277f2

Commits on Oct 9, 2018

• Improved the tokenization of round brackets, colons and semicolons of the lob corpus,

  Show description for 2b678bd
  gwenniger

  committed
  2b678bd

  Copy full SHA for 2b678bd

Commits on Oct 8, 2018

• Added brown_corpus_preprocessor.py, vocabulary_word_coverage_analysis.py and vocabulary_word_coverage_analysis.py,

  Show description for 1deaf8d
  gwenniger

  committed
  1deaf8d

  Copy full SHA for 1deaf8d

• Added a function to correction the tokenizaiton of words with an

  Show description for 32695ad
  gwenniger

  committed
  32695ad

  Copy full SHA for 32695ad

• Added a class "data_preprocessing/monolingual_data_preprocessing/vocabulary_word_coverage_analysis.py"

  Show description for e589177
  gwenniger

  committed
  e589177

  Copy full SHA for e589177

Commits on Oct 3, 2018

• Adapted the iam_dataset class to allow loading the train/dev/test

  Show description for 5736688
  gwenniger

  committed
  5736688

  Copy full SHA for 5736688

Commits on Oct 2, 2018

• Added a new class "filtered_lob_corpus_creator" replacing

  Show description for 944c349
  gwenniger

  committed
  944c349

  Copy full SHA for 944c349

• Added a new class "lob_original_preprocessor" used to generate the pre-processed

  Show description for 801a3ef
  gwenniger

  committed
  801a3ef

  Copy full SHA for 801a3ef

Commits on Sep 30, 2018

• Created a new class "data_preprocessing/monolingual_data_preprocessing/comments_cleaned_lines_extractor.py"

  Show description for 25bf97e
  gwenniger

  committed
  25bf97e

  Copy full SHA for 25bf97e

Commits on Sep 29, 2018

• Changed and renamed the method "create_train_loader" => "create_data_loader"

  Show description for b2d8b45
  gwenniger

  committed
  b2d8b45

  Copy full SHA for b2d8b45

Commits on Sep 26, 2018

• Added code for creating language models with integrated word separators

  Show description for 66060b2
  gwenniger

  committed
  66060b2

  Copy full SHA for 66060b2

Commits on Sep 18, 2018

• Added classes for creating a language model for the LOB (Lancaster-Oslo-Bergen) corpus

  Show description for 75e6ef8
  gwenniger

  committed
  75e6ef8

  Copy full SHA for 75e6ef8

Commits on Aug 27, 2018

• Added CER and WER outputs to the system and made

  Show description for 5168b47
  gwenniger

  committed
  5168b47

  Copy full SHA for 5168b47

Commits on Aug 22, 2018

• Added metrics for computing the character error rate (CER) which is based on

  Show description for 74f0db0
  gwenniger

  committed
  74f0db0

  Copy full SHA for 74f0db0

Commits on Aug 21, 2018

• Changed the method that collects the characters for IAM lines

  Show description for e4fd5a1
  gwenniger

  committed
  e4fd5a1

  Copy full SHA for e4fd5a1

• Added a version/parameterization of the MDLSTM network that

  Show description for 40ef8c8
  gwenniger

  committed
  40ef8c8

  Copy full SHA for 40ef8c8

Commits on Aug 20, 2018

• Added some more test images to test the clipping.

  Show description for d6359ad
  gwenniger

  committed
  d6359ad

  Copy full SHA for d6359ad

Pagination

Previous

Next