bgshih/crnn

Issues

• Elapsed time prection is considerably heigh

  #56 opened 7 years ago
  0

• Exception AttributeError: "'DataLoaderIter' object has no attribute 'shutdown'" in <bound method DataLoaderIter.__del__ of <torch.utils.data.dataloader.DataLoaderIter object at 0x7fe13cccd7d0>> ignored

  #55 opened 8 years ago
  0

• If the model works for variable length images, why we need to rescale images to a fixed size such as (100,32) ?

  #54 opened 7 years ago
  3

• The issue of creating a dataset with create_dataset.py with python3.5

  #53 opened 8 years ago
  12

• What version of this program does it work with?(32bit or 64bit)

  #52 opened 8 years ago
  2

• increasing variable length

  #51 opened 8 years ago
  0

• Problem with running docker

  #50 opened 8 years ago
  2

• Compilation error when execute sh build_cpp.sh

  #49 opened 8 years ago
  8

• Is the Pytorch version possible?

  #48 opened 8 years ago
  1

• are you BiRnnJoin right ?

  #47 opened 8 years ago
  0

• fblualib version

  #45 opened 6 years ago
  10

• Compilation failed, what was the environment for your program? Much thanks

  #44 opened 8 years ago
  7

• check the function: recognizeImageLexiconFree?

  #43 opened 8 years ago
  0

• batch normalized layers are not matched in config.lua and the trained model. so get the bug : ./utilities.lua:256: assertion failed!

  #42 opened 8 years ago
  1

• The image of long width has a bad result, the short one does not

  #41 opened 8 years ago
  6

• The issue of change your code?

  #40 opened 8 years ago
  0

• How to change the code to use RGB image, variable width to train ,and a new language

  #39 opened 8 years ago
  2

• Add STN to crnn

  #38 opened 8 years ago
  2

• The recognition result is very bad,such as the following iamges

  #37 opened 8 years ago
  10

• why don't answer me ?

  #36 opened 8 years ago
  0

• can anyone give train image and label for example?

  #35 opened 7 years ago
  2

• how can i use this to recognition chinese word in image?

  #34 opened 6 years ago
  7

• Pretrained model training set contains IIIT5k images too?

  #33 opened 8 years ago
  11

• Make error

  #31 opened 8 years ago
  7

• Training on manual dataset : Error : terminate called after throwing an instance of 'std::invalid_argument'

  #30 opened 8 years ago
  1

• make error

  #29 opened 8 years ago
  15

• Set Learning Rate

  #28 opened 8 years ago
  3

• How to create the user DB file?

  #27 opened 8 years ago
  0

• Confidence score of OCR

  #26 opened 8 years ago
  3

• Query - Why is it necessary to have images of fixed width

  #25 opened 7 years ago
  3

• Is there any possible to use cudnn.BLSTM to replace bidirectionalLSTM and use Baidu's warp_ctc to replace CtcCriterion?

  #24 opened 8 years ago
  6

• nothing

  #23 opened 8 years ago
  0

• Any particular reason for working on Grayscale images? Why not RGB?

  #22 opened 8 years ago
  3

• THC_LIBRARY not found

  #21 opened 8 years ago
  5

• is the ctc implementation gpu runnable?

  #20 opened 8 years ago
  1

• The Lua version you depend on is 5.1? the c++ encapsulation for lua seems differ from v5.1 to v5.2 ?

  #19 opened 8 years ago
  2

• errors with my own mdb datasets

  #18 opened 8 years ago
  5

• Can I train the Chinese character?

  #17 opened 9 years ago
  1

• Can I increase the model nClasses ?

  #16 opened 9 years ago
  2

• training errors with my own mdb datasets

  #15 opened 9 years ago
  4

• 对于CTC代码中fvars和bvars的计算问题

  #14 opened 9 years ago
  4

• Issues faced when changing Image Height from 32

  #13 opened 9 years ago
  2

• 请问代码中到底有没有限制长度

  #12 opened 9 years ago
  4

• Training of the provided model

  #11 opened 9 years ago
  1

• Question for CTC decoding

  #10 opened 9 years ago
  8

• Problems at training OMR dataset

  #9 opened 9 years ago
  10

• some errors in create_dataset.py

  #8 opened 9 years ago
  1

• Number of training steps for results from paper

  #7 opened 9 years ago
  1

• model:add(cudnn.SpatialMaxPooling(2, 2, 1, 2, 1, 0))?

  #6 opened 9 years ago
  1

• Only 62% accuracy.How to make it better?

  #5 opened 9 years ago
  17