meijieru/crnn.pytorch

Convolutional recurrent network in pytorch

Convolutional Recurrent Neural Network

This software implements the Convolutional Recurrent Neural Network (CRNN) in pytorch. Origin software could be found in crnn

Run demo

A demo program can be found in demo.py. Before running the demo, download a pretrained model from Baidu Netdisk or Dropbox. This pretrained model is converted from auther offered one by tool. Put the downloaded model file crnn.pth into directory data/. Then launch the demo by:

python demo.py

The demo reads an example image and recognizes its text content.

Example image:

Expected output: loading pretrained model from ./data/crnn.pth a-----v--a-i-l-a-bb-l-ee-- => available

Dependence

• warp_ctc_pytorch
• lmdb

Train a new model

1. Construct dataset following origin guide. If you want to train with variable length images (keep the origin ratio for example), please modify the tool/create_dataset.py and sort the image according to the text length.
2. Execute python train.py --adadelta --trainRoot {train_path} --valRoot {val_path} --cuda. Explore train.py for details.

crnn.pytorch's Issues

TypeError: 'required' is an invalid argument for positionals with crnn_main.py

Hello,
After creating my own dataset with create_dataset.py l got the following :
/train_directory which contains data-train.mdb & lock-train.mdb
/test_directory which contains data-test.mdb & lock-train.mdb

then while l run crnn_main.py l get the following error :

python2 crnn_main.py [--param val]
Traceback (most recent call last):
  File "crnn_main.py", line 18, in <module>
    parser.add_argument('/home/ahmed/Downloads/sample/output/train', required=True, help='path to dataset')
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/argparse.py", line 1276, in add_argument
    kwargs = self._get_positional_kwargs(*args, **kwargs)
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/argparse.py", line 1388, in _get_positional_kwargs
    raise TypeError(msg)
TypeError: 'required' is an invalid argument for positionals

Error related with :

parser = argparse.ArgumentParser()
parser.add_argument('/home/ahmed/Downloads/sample/output/train', required=True, help='path to dataset')
parser.add_argument('/home/ahmed/Downloads/sample/output/test', required=True, help='path to dataset')

What's wrong ?

Did you achieve the same accuracy as the original crnn?

Hi can you achieve the same accuracy as the original crnn? And what's your accuracy?

Thanks.

crnn

Does the model requires png images to perform well ?

cublas runtime error： the GPU program failed to execute

I run crnn_main.py by this command:

python crnn_main.py --trainroot="/home/wangjianbo_i/OCR/data/IIIT5K/traindatalmdb" --valroot="/home/wangjianbo_i/OCR/data/IIIT5K/testdatalmdb" --cuda --alphabet="0123456789abcdefghijklmnopkrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ,;:" --imgH=32

And the error message shows below. I tried to run it on the company's server to avoid the environment error, but get the same result. How to solve this problem? :-D

why the size of images in test are (100,32) instead of (128,32)

• root=opt.valroot, transform=dataset.resizeNormalize((128, 32)))

• root=opt.valroot, transform=dataset.resizeNormalize((100, 32)))

when training, we use (128,32), why (100,32) use in test?

text = torch.IntTensor(opt.batchSize * 5)????

the main function line 91:text = torch.IntTensor(opt.batchSize * 5),can you explain what this line meaning?

AttributeError: module 'utils' has no attribute 'data_parallel'

Hi! When I was trying your demo program, I got this error:

loading pretrained model from ./data/crnn.pth
Traceback (most recent call last):
File "demo.py", line 27, in
preds = model(image)
File "/home/rohit/.local/lib/python3.5/site-packages/torch/nn/modules/module.py", line 206, in call
result = self.forward(*input, **kwargs)
File "/home/rohit/Documents/gitprojects/crnn.pytorch/models/crnn.py", line 78, in forward
conv = utils.data_parallel(self.cnn, input, self.ngpu)
AttributeError: module 'utils' has no attribute 'data_parallel'

Could you please tell me how I can fix this?

Confusion matrix for CRNN

Hello,

l'm thinking about making confusion matrix for CRNN. However the output classes are variable sequences and not as standard output classes as 0 1 2 .. or A B C .
Any idea of how making confusion matrix for CRNN.

Thank you

AssertionError: the height of conv must be 1

I keep getting this error on running crnn_main.py.
Traceback (most recent call last): File "crnn_main.py", line 200, in <module> cost = trainBatch(crnn, criterion, optimizer) File "crnn_main.py", line 183, in trainBatch preds = crnn(image) File "/home/ubuntu/anaconda/lib/python2.7/site-packages/torch/nn/modules/module.py", line 206, in __call__ result = self.forward(*input, **kwargs) File "/home/ubuntu/crnn.pytorch/models/crnn.py", line 80, in forward assert h == 1, "the height of conv must be 1" AssertionError: the height of conv must be 1

can you give train image and label for example?

I create the lmdb data, and the model is worse。I think the ground truth I set is wrong.

training data format?

is the img name is the label?can you show me the training data format

how can l fine tune the model ?

training on custom dataset

@meijieru
Thanks for this package. I would like to know that I want to train it on a different dataset, but when I use /tools/create_dataset.py, I get nothing. So please help me in this matter or suggest me any other method to do so.

nn.parallel.data_parallel(self.cnn, input, gpu_ids) giving error

The following error I am getting while runnunig demo.py for crnn inference.can you please help me:

loading pretrained model from ./data/crnn.pth
Traceback (most recent call last):
  File "demo.py", line 27, in <module>
    preds = model(image)
  File "/home/anirban/python/local/lib/python2.7/site-packages/torch/nn/modules/module.py", line 206, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/anirban/crnn_inference/models/crnn.py", line 86, in forward
    conv = nn.parallel.data_parallel(self.cnn, input, gpu_ids)
  File "/home/anirban/python/local/lib/python2.7/site-packages/torch/nn/parallel/data_parallel.py", line 96, in data_parallel
    output_device = device_ids[0]
TypeError: 'NoneType' object has no attribute '__getitem__'

when using Multi-GPU training , get "all tensors must be on devices[0]"

what should i do?

RuntimeError when launching demo.py

When i try to launch demo.py i get this error:
Traceback (most recent call last):
File "demo.py", line 30, in
preds = preds.squeeze(2)
File "/home/ubuntu-andrea/pytorchGPU3/lib/python3.5/site-packages/torch/autograd/variable.py", line 717, in squeeze
return Squeeze.apply(self, dim)
File "/home/ubuntu-andrea/pytorchGPU3/lib/python3.5/site-packages/torch/autograd/_functions/tensor.py", line 375, in forward
result = input.squeeze(dim)
RuntimeError: dimension out of range (expected to be in range of [-2, 1], but got 2)

What could be the problem? @meijieru

using gpu error:"'module' object has no attribute 'gpu_ctc'"

run crnn_main.py using --cuda,and then error.
Error content：loss_func = warp_ctc.gpu_ctc if is_cuda else warp_ctc.cpu_ctc
AttributeError:'module' object has no attribute 'gpu_ctc'.
How to deal with?

Why preds variable contains negative values?

I have looked at the values of variable preds after executing "preds = model(image)" in demo.py. The values are the following:
Variable containing:
(0 ,.,.) =
-106.3455 -115.3943 -114.5584 ... -115.6788 -110.2145 -112.2794

(1 ,.,.) =
-67.3953 -92.3248 -92.7227 ... -88.7459 -81.5368 -88.8212

(2 ,.,.) =
-56.8008 -89.8197 -92.8852 ... -85.0180 -77.4713 -85.3732
...

(35,.,.) =
-38.8606 -79.6700 -81.3100 ... -71.8229 -57.3992 -68.8093

(36,.,.) =
-39.6410 -75.7699 -75.7648 ... -70.3662 -55.7602 -68.3655

(37,.,.) =
-45.2289 -77.6819 -77.0921 ... -73.9425 -59.1527 -70.6211
[torch.cuda.FloatTensor of size 38x1x37 (GPU 0)]

For my concerning this values should represents a sequence of probabilities over the classes. So i'm wondering why they are negative values such the ones reported here? @meijieru

the net doesn't work,when I run the net, it stopped by the flowing informations

[dutongchun@cpu0 crnn.pytorch]$ python crnn_main.py  --trainroot /home/dutongchun/songwendong/xxx/train   --valroot /home/dutongchun/songwendong/xxx/validate/ --batchSize 16 --workers 1  --cuda 
Namespace(adadelta=False, adam=False, alphabet='0123456789abcdefghijklmnopqrstuvwxyz', batchSize=16, beta1=0.5, crnn='', cuda=True, displayInterval=500, experiment=None, imgH=32, imgW=100, keep_ratio=False, lr=0.01, n_test_disp=10, ngpu=1, nh=256, niter=25, random_sample=False, saveInterval=500, trainroot='/home/dutongchun/songwendong/xxx/train', valInterval=500, valroot='/home/dutongchun/songwendong/xxx/validate/', workers=1)
Random Seed:  2008
CRNN (
  (cnn): Sequential (
    (conv0): Conv2d(1, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (relu0): ReLU (inplace)
    (pooling0): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (relu1): ReLU (inplace)
    (pooling1): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
    (conv2): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (batchnorm2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True)
    (relu2): ReLU (inplace)
    (conv3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (relu3): ReLU (inplace)
    (pooling2): MaxPool2d (size=(2, 2), stride=(2, 1), dilation=(1, 1))
    (conv4): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (batchnorm4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True)
    (relu4): ReLU (inplace)
    (conv5): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (relu5): ReLU (inplace)
    (pooling3): MaxPool2d (size=(2, 2), stride=(2, 1), dilation=(1, 1))
    (conv6): Conv2d(512, 512, kernel_size=(2, 2), stride=(1, 1))
    (batchnorm6): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True)
    (relu6): ReLU (inplace)
  )
  (rnn): Sequential (
    (0): BidirectionalLSTM (
      (rnn): LSTM(512, 256, bidirectional=True)
      (embedding): Linear (512 -> 256)
    )
    (1): BidirectionalLSTM (
      (rnn): LSTM(256, 256, bidirectional=True)
      (embedding): Linear (512 -> 37)
    )
  )
)
/home/dutongchun/songwendong/crnn.pytorch/dataset.py:95: UserWarning: torch.range is deprecated in favor of torch.arange and will be removed in 0.3. Note that arange generates values in [start; end), not [start; end].
  batch_index = random_start + torch.range(0, self.batch_size - 1)
[dutongchun@cpu0 crnn.pytorch]$ 

RuntimeError: cuda runtime error (2) : out of memory at /py/conda-bld/pytorch_1493676237139/work/torch/lib/THC/generic/THCStorage.cu:66

Hello,
When l run crnn_main.py with more than 15000 training exmples l get the following error :
and also when l run it with less than 10000 examples but l execute 3 or more times :

THCudaCheck FAIL file=/py/conda-bld/pytorch_1493676237139/work/torch/lib/THC/generic/THCStorage.cu line=66 error=2 : out of memory
Traceback (most recent call last):
  File "crnn_main.py", line 222, in <module>
    cost = trainBatch(crnn, criterion, optimizer)
  File "crnn_main.py", line 207, in trainBatch
    cost.backward()
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/site-packages/torch/autograd/variable.py", line 146, in backward
    self._execution_engine.run_backward((self,), (gradient,), retain_variables)
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/site-packages/torch/nn/_functions/thnn/auto.py", line 171, in backward
    grad_input = input.new().resize_as_(input)
RuntimeError: cuda runtime error (2) : out of memory at /py/conda-bld/pytorch_1493676237139/work/torch/lib/THC/generic/THCStorage.cu:66


How can l circumvent that  ?

Thank you

Error while converting crnn_demo_model.t7 into a .pth model

@meijieru
I am trying to convert the crnn_demo_model.t7 into an equivalent .pth model using the script convert_t7.py but I am getting the following error.

File "tool/convert_t7.py", line 167, in
torch_to_pytorch(py_model, args.model_file, args.output)
File "tool/convert_t7.py", line 130, in torch_to_pytorch
if layer_map[t7_name] != py_name:
KeyError: 'b'

Please help me to rectify this issue.

Probelms when fine tuning the models . RuntimeError: inconsistent tensor size

Hello,

l stuck with fine tuning.

1)First of all to fine tune the model you have to set --nh="256" otherwise it will not work, you'll get this error

(
loading pretrained model from /home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth
Traceback (most recent call last):
File "crnn_main.py", line 98, in
crnn.load_state_dict(torch.load(opt.crnn))
File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/site-packages/torch/nn/modules/module.py", line 335, in load_state_dict
own_state[name].copy_(param)
RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51
)

because the pretrained model --nh="256" and not 100 as it is set in the default model. But when fine tuning obviously we can change the parameter, so l find it strange that it doesn't work

2. tried different configurations while fine tuning
   the length of the alphabet , nb_classes= 37 '0123456789abcdefghijklmnopqrstuvwxyz' by default

l tried the following :
A) add one letter, let's say Z or another char , . /
'0123456789abcdefghijklmnopqrstuvwxyzZ'
l got the same error

RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

B) l removed one char and add another remove z and add / '0123456789abcdefghijklmnopqrstuvwxy/'

l get the same error

RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

C) l set alphabet only to digits '0123456789'

the same error

RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

3. train a new model with a varibale length alphabet and number of -nh it works perfectly

Have you any idea for solving the problem of fine tuning to make a variable length of alphabet and the architecture ?
Thanks a lot

Training got stuck with cuda

Hi, when I train my own model with cpu, it goes normal and prints out loss. However if I use GPU, the system just get stuck there, after printing out model configuration it does not print anything. Do you know what might be the problem?

AttributeError: module 'utils' has no attribute 'data_parallel' when running python demo.py

Hello,
l'm getting this error when running python demo.py.
What's wrong with my the code. It seems that crnn.py don't import the utils which is in crnn.pytorch/models/ and import the one which is located in crnn.pytorch/

 python3.5 demo.py 
loading pretrained model from ./data/crnn.pth
Traceback (most recent call last):
  File "demo.py", line 27, in <module>
    preds = model(image)
  File "/home/ahmed/anaconda3/envs/my_env/lib/python3.5/site-packages/torch/nn/modules/module.py", line 206, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/ahmed/Downloads/crnn.pytorch-master/models/crnn.py", line 78, in forward
    conv = utils.data_parallel(self.cnn, input, self.ngpu)
AttributeError: module 'utils' has no attribute 'data_parallel'

About the process of training a new model

Hi meijieru,

First of all, thanks so much for the package, it works nicely and is easy to install and test.

I am really interested in your package and I am trying to go deeper by redoing the whole process starting from training the model. I am quite a beginner in this field, so I have some question as follows,

1. If I understand correctly, to train a model, I need to prepare a training dataset (a set of images of words and their corresponding labels) and pass it to "create_dataset.py" of the original "crnn" package. Then I have to use "convert_dataset.py" file in the "tool" folder to sort the training files according to their width. Finally use "crnn_main.py" to train the model. Is that correct?

2. Is possible, is there anywhere that I can download the training data to reproduce the exactly same (or similar) result (final trained model) as you got in the end?

I knew these questions might sound silly to you, but your comments are greatly appreciated. I think it is also helpful to others who are new to this field. Many thanks.

Su

Issues related to dataset.resizeNormalize((100, 32))

Hello,
l noticed that all the images are constrained to be in the same size.
1- transformer = dataset.resizeNormalize((100, 32)) in demo.py
img_path = '/demo.png' of the sequence "available" has a dimension of (184,72), when we rescale it to (100,32) the amount of lost information is negligible. But when we have images of variable sizes far from 100 and 32 this will be problematic. In this case how did you solve the problem.

2- the same question than the one asked in (1-) transform=dataset.resizeNormalize((128, 32))) in crnn_main.py. Why we should resize all the images to (128,32) ?
for the height=32 l do agree because it allows to make difference for example between 'i' and 'l' so the height is important to discriminate between characters, but for the width=128 l don't understand why it should be!

3- Does the parameters transform=dataset.resizeNormalize((128, 32))) in crnn_main.py should remain statics to (128,32) to be able to use the architecture or can be changed to the dimensions which fit our data ?

when using Multi-GPU training , get "ValueError: result of slicing is an empty tensor"

My OS is CentOS Linux release 7.2.1511 (Core), x86_64.
My server has four Tesla P40 GPU cards.

The command is:
python crnn_main.py --trainroot /home/xuliang/CRNN_org/crnn/tool/synth90k_train_sort --valroot /home/xuliang/CRNN_org/crnn/tool/synth90k_val_sort --cuda --ngpu 4 --adadelta --keep_ratio --random_sample
And error message is:
Namespace(adadelta=True, adam=False, alphabet='0123456789abcdefghijklmnopqrstuvwxyz', batchSize=64, beta1=0.5, crnn='', cuda=True, displayInterval=500, experiment=None, imgH=32, imgW=100, keep_ratio=True, lr=0.01, n_test_disp=10, ngpu=4, nh=256, niter=25, random_sample=True, saveInterval=500, trainroot='/home/xuliang/CRNN_org/crnn/tool/synth90k_train_sort', valInterval=500, valroot='/home/xuliang/CRNN_org/crnn/tool/synth90k_val_sort', workers=2)
Random Seed: 7716
CRNN (
(cnn): Sequential (
(conv0): Conv2d(1, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(relu0): ReLU (inplace)
(pooling0): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
(conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(relu1): ReLU (inplace)
(pooling1): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))
(conv2): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(batchnorm2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True)
(relu2): ReLU (inplace)
(conv3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(relu3): ReLU (inplace)
(pooling2): MaxPool2d (size=(2, 2), stride=(2, 1), dilation=(1, 1))
(conv4): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(batchnorm4): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True)
(relu4): ReLU (inplace)
(conv5): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(relu5): ReLU (inplace)
(pooling3): MaxPool2d (size=(2, 2), stride=(2, 1), dilation=(1, 1))
(conv6): Conv2d(512, 512, kernel_size=(2, 2), stride=(1, 1))
(batchnorm6): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True)
(relu6): ReLU (inplace)
)
(rnn): Sequential (
(0): BidirectionalLSTM (
(rnn): LSTM(512, 256, bidirectional=True)
(embedding): Linear (512 -> 256)
)
(1): BidirectionalLSTM (
(rnn): LSTM(256, 256, bidirectional=True)
(embedding): Linear (512 -> 37)
)
)
)
[0/25][500/112885] Loss: 16.009935
Start val
Traceback (most recent call last):
File "crnn_main.py", line 207, in 
val(crnn, test_dataset, criterion)
File "crnn_main.py", line 158, in val
sim_preds = converter.decode(preds.data, preds_size.data, raw=False)
File "/home/xuliang/CRNN_pytorch_v2/crnn.pytorch/utils.py", line 51, in decode
t[index:index + l], torch.IntTensor([l]), raw=raw))
ValueError: result of slicing is an empty tensor
Thanks for your help.

can not find the training data directory

GPU

Training loss and val loss decrease normally, but accuracy is always 0

I use the icdar dataset to train crnn(pytorch), here is the problem:
Both training loss and test loss decrease normally, but the network can not predict anything!
Dose anybody encounter similar problem?
Here is an screenshot of my result:

Start val
-------------------------- => , gt: Oleg
-------------------------- => , gt: P
-------------------------- => , gt: CASTLE
-------------------------- => , gt: HIGH
-------------------------- => , gt: GYM
-------------------------- => , gt: FOR
-------------------------- => , gt: COOLING
-------------------------- => , gt: only
-------------------------- => , gt: Refill
-------------------------- => , gt: GRADUATE
Test loss: 19.685625, accuray: 0.000000
[7/25][60/60] Loss: 19.398106
[8/25][20/60] Loss: 19.721890
[8/25][40/60] Loss: 19.135641
Start val
-------------------------- => , gt: Hypertension:
-------------------------- => , gt: ntl:
-------------------------- => , gt: ANCIENT
-------------------------- => , gt: PanaSyncE70
-------------------------- => , gt: Specialists
-------------------------- => , gt: 2N
-------------------------- => , gt: EU-funded
-------------------------- => , gt: Dez.
-------------------------- => , gt: Guidex
-------------------------- => , gt: TAXI
Test loss: 19.580940, accuray: 0.000000
[8/25][60/60] Loss: 20.216869
[9/25][20/60] Loss: 18.244850
[9/25][40/60] Loss: 19.222029
Start val
-------------------------- => , gt: Connells
-------------------------- => , gt: 2010
-------------------------- => , gt: &
-------------------------- => , gt: GENTLEMEN
-------------------------- => , gt: WELCOME
-------------------------- => , gt: STUDIO
-------------------------- => , gt: DOLLAR
-------------------------- => , gt: SPENCER
-------------------------- => , gt: Systems
-------------------------- => , gt: NIGHT
Test loss: 19.106692, accuray: 0.000000
[9/25][60/60] Loss: 18.840774

TypeError: __init__() takes from 1 to 4 positional arguments but 9 were given convert_dataset.py

Hello,
to run convert_dataset.py l give the path to the output of create_dataset.py which are data.mdb and lock.mdb

When l execute convert_dataset.py

if __name__ == "__main__":
    convert('/home/ahmed/Downloads/sample/data/training', '/home/ahmed/Downloads/sample/data/training-ordered')

l get the following error :

Traceback (most recent call last):
  File "/home/ahmed/Downloads/crnn.pytorch-master/tool/convert_dataset.py", line 55, in <module>
    convert('/home/ahmed/Downloads/sample/data/training', '/home/ahmed/Downloads/sample/data/data-ordered')
  File "/home/ahmed/Downloads/crnn.pytorch-master/tool/convert_dataset.py", line 17, in convert
    originDataset = dataset.lmdbDataset(originPath, 'abc', *args)
TypeError: __init__() takes from 1 to 4 positional arguments but 9 were given

Issue while training

Start val
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: accredits
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: cabral
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: esters
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: indulgent
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: typologies
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: bulwarks
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: insensitivity
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: fiche
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: elephantiasis
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: niamey
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: divorces
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: uncanny
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: nameable
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: colds
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: hydrofoils
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: eccentrically
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: strictly
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: misappropriate
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: televisions
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: turgid
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: undefined
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: margaritas
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: notifying
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: amniocentesis
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: lands
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: cunard
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: camshaft
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: taps
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: mediated
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: flexes
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: plumier
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: extradite
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: paramountcy
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: korans
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: compel
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: homburg
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: noncommunicable
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: fireguard
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: servants
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: dinnertime
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: elopes
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: innervate
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: whys
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: unhurt
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: screechier
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: chops
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: saprophytes
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: cannily
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: laborers
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: inns
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: loused
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: agrees
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: gusty
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: prizewinners
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: cremation
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: notoriously
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: awesome
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: musicals
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: cuddliest
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: unsupervised
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: villus
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: daters
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: acheson
qqqqqqqqqqqqqqqqqqqqqqqqqq => q , gt: unmentioned
Test loss: inf, accuray: 0.000000

I get this output. It doesn't change. The loss is infinite and the accuracy is zero. Is there some parameter I should tweak?

Issue with LMDB file while learning

Hello,

I've created a LMDB file with Caffe but i have some issues while reading it with six.BytesIO() so i can't get the names of my images. It tells me that my images are corrupted (but the name in the lmdb is right).
Is there any other methods to create a LMDB file (without Caffe maybe) or other suggestions ?

I'm working on Ubuntu 16.04 with python3.4

Thank you in advance.
Best regards

Elapsed time prection on the first sequence takes considerably more time

Hello,
l've tested the pre-trained model in (demo.py) with IIIT5K test dataset as mentioned in the paper as follow :
l took 18 images. l wanted to know how much time it takes for each sequence image prediction.

import torch
from torch.autograd import Variable
import utils
import dataset
from PIL import Image

import models.crnn as crnn
from datetime import datetime
import glob
import os

test_path='/data/IIITS5K/'

alphabet = '0123456789abcdefghijklmnopqrstuvwxyz'


model = crnn.CRNN(32, 1, 37, 256, 1).cuda()
print('loading pretrained model from %s' % model_path)
model.load_state_dict(torch.load(model_path))

converter = utils.strLabelConverter(alphabet)

transformer = dataset.resizeNormalize((100, 32))

os.chdir(test_path)
images_names=glob.glob("*.png")
i=0

p=0
for img in images_names:
    image = Image.open(img).convert('L')

   
    a = datetime.now()

    image = transformer(image).cuda()
    image = image.view(1, *image.size())
    image = Variable(image)

    model.eval()
    preds = model(image)

    _, preds = preds.max(2)
    preds = preds.squeeze(2)
    preds = preds.transpose(1, 0).contiguous().view(-1)

    preds_size = Variable(torch.IntTensor([preds.size(0)]))
    raw_pred = converter.decode(preds.data, preds_size.data, raw=True)
    sim_pred = converter.decode(preds.data, preds_size.data, raw=False)
    b = datetime.now()
    c = b - a
    print('%-20s => %-20s' % (raw_pred, sim_pred),img)
    # print(end - start)
    print(c.total_seconds())
    if (sim_pred+'.png') == img:
        p +=1

    i +=1

print(p, " sur ", i , " true positive")

l noticed that always the first sequence image prediction takes from 0.453489 to 0.503489 seconds
then all the remaining images take a negligible time 0.008099 for instance. It's a little bit strange.
I'm wondering why the first sequence image takes considerably more time than the remaining images. even if we change the order of the image to process first.

crnn.pytorch-master/demo.py
loading pretrained model from /home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth
8----8--3---9--6--6--0---- => 8839660              8839660.png
0.503489
9----------3------0------- => 930                  930.png
0.009706
00-------8-----0----0----- => 0800                 0800.png
0.008247
0-----1--9----9---0----1-- => 019901               019901.png
0.008314
g---------5------m-------- => g5m                  95m.png
0.008099
1--------44-------9------- => 149                  149.png
0.008319
0------11--9---2----2----- => 01922                01922.png
0.008478
7-----8---3-1--4--2--3---- => 7831423              7831423.png
0.008663
77---------4--------1----- => 741                  741.png
0.008374
2-------0------1----0----- => 2010                 2010.png
0.008169
22------------88---------- => 28                   28.png
0.008131
t-------------o----------- => to                   to.png
0.008403
8---------------0--------- => 80                   80.png
0.008148
2----2--4---3--5--0---2--- => 2243502              2243502.png
0.008258
22--------0----0----3----- => 2003                 2003.png
0.008829
1--------8-----7----7----- => 1877                 1877.png
0.008178
2----------------55------- => 25                   25.png
0.0081
5------6----1---6----1---- => 56161                56161.png
0.009257

Why only use lowercase letters?

I mean ICDAR2013 evaluation includes uppercase letters right ?

License?

Is it open source? Which license?

TypeError: Won't implicitly convert Unicode to bytes; use .encode() : Error related to create a dataset with create_dataset.py :

Hello,
l'm trying to create my own dataset to train my model from scratch. I'm working with python3.5.2 .While l'm creating a dataset l encountered the following problems.
1)

with open(imagePath, 'r') as f:
     imageBin = f.read()

returns the following error :

codecs.py", line 321, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

l fixed it as follow :

with open(imagePath, 'rb') as f:
       imageBin = f.read()

2)for python3

for i in xrange(nSamples): should be replaced by for i in range(nSamples):
and
for k, v in cache.iteritems(): should be replaced by for k, v in cache.items():

3. However, l got the following error at this line :

txn.put(k, v) indef writeCache(env, cache): function

line 86, in createDataset
   writeCache(env, cache)
 File "/home/ahmed/Downloads/crnn.pytorch-master/data-processing.py", line 46, in writeCache
   txn.put(k, v)
TypeError: Won't implicitly convert Unicode to bytes; use .encode()

Here is my whole code :

import os
import lmdb # install lmdb by "pip install lmdb"
import cv2
import numpy as np
import glob

real_path='/home/ahmed/Downloads/sample/train/'
path='/home/ahmed/Downloads/sample/'
path_train='train/'

path_output='/home/ahmed/Downloads/sample/output/'



os.chdir(path+path_train)
images_train = glob.glob("*.jpg")
left_train,labels_train,right_train = list(zip(*[os.path.splitext(x)[0].split('_')
                                         for x in images_train]))



def checkImageIsValid(imageBin):
    if imageBin is None:
        return False
    imageBuf = np.fromstring(imageBin, dtype=np.uint8)
    img = cv2.imdecode(imageBuf, cv2.IMREAD_GRAYSCALE)
    imgH, imgW = img.shape[0], img.shape[1]
    if imgH * imgW == 0:
        return False
    return True


def writeCache(env, cache):
    with env.begin(write=True) as txn:
        for k, v in cache.items():
            txn.put(k, v)


def createDataset(outputPath,images_train, labels_train, lexiconList=None, checkValid=True):
    """
    Create LMDB dataset for CRNN training.
    ARGS:
        outputPath    : LMDB output path
        imagePathList : list of image path
        labelList     : list of corresponding groundtruth texts
        lexiconList   : (optional) list of lexicon lists
        checkValid    : if true, check the validity of every image
    """
    assert(len(images_train) == len(labels_train))
    nSamples = len(images_train)

    env = lmdb.open(path_output, map_size=1099511627776)
    cache = {}
    cnt = 1
    for i in range(nSamples):
        imagePath = images_train[i]
        label = labels_train[i]
        if not os.path.exists(real_path+imagePath):
            print('%s does not exist' % imagePath)
            continue
        with open(real_path+imagePath, 'rb') as f:
            imageBin = f.read()
        if checkValid:
            if not checkImageIsValid(imageBin):
                print('%s is not a valid image' % imagePath)
                continue

        imageKey = 'image-%09d' % cnt
        labelKey = 'label-%09d' % cnt
        cache[imageKey] = imageBin
        cache[labelKey] = label
        if lexiconList:
            lexiconKey = 'lexicon-%09d' % cnt
            cache[lexiconKey] = ' '.join(lexiconList[i])
        if cnt % 1000 == 0:
            writeCache(env, cache)
            cache = {}
            print('Written %d / %d' % (cnt, nSamples))
        cnt += 1
    nSamples = cnt-1
    cache['num-samples'] = str(nSamples)
    writeCache(env, cache)
    print('Created dataset with %d samples' % nSamples)


if __name__ == '__main__':
    createDataset(path_output,images_train, labels_train)

run python demo.py , failed reporting error ：dimension out of range (expected to be in range of [-2, 1], but got 2)

When I run "python demo.py", but failed, met these error:

loading pretrained model from ./data/crnn.pth
Traceback (most recent call last):
File "demo.py", line 30, in
preds = preds.squeeze(2)
File "/usr/local/lib/python2.7/dist-packages/torch/autograd/variable.py", line 704, in squeeze
return Squeeze.apply(self, dim)
File "/usr/local/lib/python2.7/dist-packages/torch/autograd/_functions/tensor.py", line 341, in forward
result = input.squeeze(dim)
RuntimeError: dimension out of range (expected to be in range of [-2, 1], but got 2)

could you help me how to solve this problem?

Loss going to infinity after one minibatch

I made a lmdb dataset from synth90k and the crnn is able to run. But the loss goes to infinity immediately. How can I fix this?
(I also changed imgH to 32)

sixlmdp_train_data_ordered module is missing

Hi Jieru,

Thanks for your previous answer. I managed to create my own dataset. However, when I ran the "crnn_main.py" file with args specifying the training and testing dataset (with everything else the same as default value), it gave me errors as follows:

Traceback (most recent call last):
  File "/home/su/work_space/pycharm-community-2017.1/helpers/pydev/pydevd.py", line 1578, in <module>
    globals = debugger.run(setup['file'], None, None, is_module)
  File "/home/su/work_space/pycharm-community-2017.1/helpers/pydev/pydevd.py", line 1015, in run
    pydev_imports.execfile(file, globals, locals)  # execute the script
  File "/home/su/work_space/MOCR/crnn.pytorch/crnn_main.py", line 13, in <module>
    import dataset
  File "/home/su/work_space/MOCR/crnn.pytorch/dataset.py", line 10, in <module>
    import sixlmdp_train_data_ordered
ImportError: No module named sixlmdp_train_data_ordered

Does it mean the python module "sixlmdp_train_data_ordered" is somehow missing? Is it a custom module defined by yourself? I made a bit of search, but I can't find it anywhere online. Many thanks.

Cheers,

Su

how to run with cpu

i want some data about running with CPU, how can i turn off CUDA ?

Fine-tuning the model : RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

Hello,
l want to fine-tune the model on my data. so l added this instruction in my command to load the pretrained model

--crnn="/home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth"

as follow :

python2 crnn_main.py --trainroot="/home/ahmed/Downloads/mnt/data/train" --valroot="/home/ahmed/Downloads/mnt/data/valid" --imgH=32 --cuda --crnn="/home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth" --adadelta --experiment="/home/ahmed/Downloads/training_data/output"

However l get the following error :

Namespace(Diters=5, adadelta=True, adam=False, alphabet='0123456789abcdefghijklmnopqrstuvwxyz', batchSize=64, beta1=0.5, crnn='/home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth', cuda=True, displayInterval=1, experiment='/home/ahmed/Downloads/training_data/output', imgH=32, keep_ratio=False, lr=1, n_test_disp=1, ngpu=1, nh=100, niter=25, random_sample=False, saveInterval=1, trainroot='/home/ahmed/Downloads/mnt/data/train', valInterval=1, valroot='/home/ahmed/Downloads/mnt/data/valid', workers=2)
Random Seed:  5051
loading pretrained model from /home/ahmed/Downloads/crnn.pytorch-master/data/crnn.pth
Traceback (most recent call last):
  File "crnn_main.py", line 97, in <module>
    crnn.load_state_dict(torch.load(opt.crnn))
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/site-packages/torch/nn/modules/module.py", line 335, in load_state_dict
    own_state[name].copy_(param)
RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

What's wrong ?
l don't understand the error :

RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51
Every comment would be appreciable
Thank you

torch.FloatTensor(opt.batchSize, 3, opt.imgH, opt.imgH) in crnn_main.py

Hello,
l stack at the following line code in crnn_main.py

1. in line 100
   image = torch.FloatTensor(opt.batchSize, 3, opt.imgH, opt.imgH)
   we are are supposed to have imgH, imgW or you rescale the image to be symmetric 32 * 32 ?

2. line 188
   cost.backward()
   How it backward the cost ?

3. in line 101
   text = torch.IntTensor(opt.batchSize * 5)
   Why is it batchSize*5? is 5 a static parameter or can be changed ?

AttributeError: 'lmdbDataset' object has no attribute 'getLabel' in convert_dataset.py. getLabel Missing

Hello,
when l run convert_dataset.py l get the following error :

python2 convert_dataset.py 
Origin dataset has 23920 samples
Traceback (most recent call last):
  File "convert_dataset.py", line 57, in <module>
    convert('/home/ahmed/Downloads/sample/output', '/home/ahmed/Downloads/sample/o_train')
  File "convert_dataset.py", line 25, in convert
    label = originDataset.getLabel(i + 1)
AttributeError: 'lmdbDataset' object has no attribute 'getLabel'

indeed, in dataset.py there is no getLabel
How to fix that ?

How can l get the probability of the sequence outputted by CRNN ?

Hello,

l'm wondering whether the CRNN is able to output also the probability of each sequence

from example :

--h-e--ll-oo- => 'hello' with a probability= 0.89
for instance
how can l get that ?

cnn height equals 1 assertion error

Hi Jieru,

Sorry for the previous question.
I cloned the code again and this time I'm getting cnn height equals 1 assertion error in crnn.py.

    def forward(self, input):
        # conv features
        conv = utils.data_parallel(self.cnn, input, self.ngpu)
        b, c, h, w = conv.size()
        assert h == 1, "the height of conv must be 1"

In the code, the input is of dimension 64x1x64x128 ([torch.cuda.FloatTensor of size 64x1x64x128 (GPU 0)]), and the conv size has become 64x512x3x33 (b=64, c=512, h=3, w=33). Could you please give me any hint on how to resolve it? My training data is sub-sample from synth90k, so if I am correct, the image should be gray scale. Many thanks.

Cheers,

Su

make[1]: *** [CMakeFiles/test_cpu.dir/all] Error 2 when building warp_ctc_pytorch

Hello ,
l tried to install the fast CTC parallel library from this link https://github.com/SeanNaren/warp-ctc as follow :

git clone https://github.com/baidu-research/warp-ctc.git
cd warp-ctc
mkdir build 
cd build 
cmake ../

-- The C compiler identification is GNU 4.8.5
-- The CXX compiler identification is GNU 4.8.5
-- Check for working C compiler: /root/anaconda3/bin/cc
-- Check for working C compiler: /root/anaconda3/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /root/anaconda3/bin/c++
-- Check for working CXX compiler: /root/anaconda3/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
**CMake Warning at /usr/share/cmake-3.5/Modules/FindCUDA.cmake:771 (message):
  Expecting to find librt for libcudart_static, but didn't find it.
Call Stack (most recent call first):
  CMakeLists.txt:20 (FIND_PACKAGE)**


-- Found CUDA: /usr/local/cuda (found suitable version "8.0", minimum required is "6.5") 
-- cuda found TRUE
-- Found Torch7 in /home/ahmed/torch/install
-- Torch found /home/ahmed/torch/install/share/cmake/torch
-- Building shared library with GPU support
-- Building Torch Bindings with GPU support
-- Configuring done
-- Generating done
-- Build files have been written to: /home/ahmed/warp-ctc/build

After running

make

l got the following error :

make
[ 10%] Building NVCC (Device) object CMakeFiles/warpctc.dir/src/warpctc_generated_reduce.cu.o
[ 20%] Building NVCC (Device) object CMakeFiles/warpctc.dir/src/warpctc_generated_ctc_entrypoint.cu.o
Scanning dependencies of target warpctc
[ 30%] Linking CXX shared library libwarpctc.so
[ 30%] Built target warpctc
Scanning dependencies of target test_cpu
[ 40%] Building CXX object CMakeFiles/test_cpu.dir/tests/test_cpu.cpp.o
[ 50%] Linking CXX executable test_cpu
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::~basic_string()@GLIBCXX_3.4.21'
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_append(char const*, unsigned long)@GLIBCXX_3.4.21'
/home/ahmed/torch/install/lib/libTH.so.0: undefined reference to `GOMP_parallel@GOMP_4.0'
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::random_device::_M_init(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)@GLIBCXX_3.4.21'
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::runtime_error::runtime_error(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)@GLIBCXX_3.4.21'
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_replace(unsigned long, unsigned long, char const*, unsigned long)@GLIBCXX_3.4.21'
/home/ahmed/torch/install/lib/libTHC.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_create(unsigned long&, unsigned long)@GLIBCXX_3.4.21'
collect2: error: ld returned 1 exit status
CMakeFiles/test_cpu.dir/build.make:99: recipe for target 'test_cpu' failed
make[2]: *** [test_cpu] Error 1
CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/test_cpu.dir/all' failed
make[1]: *** [CMakeFiles/test_cpu.dir/all] Error 2
Makefile:127: recipe for target 'all' failed
make: *** [all] Error 2

what's wrong l don't understand the error ?

raise StopIteration error in crrn_main.py DUE to max_iter

Hello,

in crnn_main.py line 125 the function def val(net, dataset, criterion, max_iter=100): takes a a parameter max_iter which by default is set to 100 . However, it doesn't work for any dataset ( but in the article the max iteration is between 200k and 500k). For huge dataset max_iter=100 works but for small one :
you should reduce the value manually.
For instance l tried two small dataset the first worked with max_iter=49 and another one with max_iter=22 and another one with max_iter=5
otherwise l get the following error

[0/35][1/13] Loss: 65.981506
Start val
Traceback (most recent call last):
  File "crnn_main.py", line 216, in <module>
    val(crnn, test_dataset, criterion)
  File "crnn_main.py", line 144, in val
    data = val_iter.next()
  File "/home/ahmed/anaconda3/envs/cv/lib/python2.7/site-packages/torch/utils/data/dataloader.py", line 202, in __next__
    raise StopIteration
StopIteration

l reached to set manually max_iter as follow :
putting print(i) in line 141

   for i in range(max_iter):
            print(i)
           data = val_iter.next()
  

to see after which iteration it returns raise StopIteration error

How to solve that ?

Thank you

error: invalid argument '-std=c++11' not allowed with 'C/ObjC'

I install with SeanNaren/warp-ctc on my macbookpro. How to solve it ?

➜  pytorch_binding git:(zxdev_pytorch) python setup.py install
CUDA_HOME not found in the environment so building without GPU support. To build with GPU support please define the CUDA_HOME environment variable. This should be a path which contains include/cuda.h
generating build/_warp_ctc.c
regenerated: 'build/_warp_ctc.c'
running install
running build
running build_py
creating build/lib.macosx-10.11-x86_64-2.7
creating build/lib.macosx-10.11-x86_64-2.7/warpctc_pytorch
copying warpctc_pytorch/__init__.py -> build/lib.macosx-10.11-x86_64-2.7/warpctc_pytorch
running build_ext
building 'warpctc_pytorch._warp_ctc' extension
creating build/temp.macosx-10.11-x86_64-2.7
creating build/temp.macosx-10.11-x86_64-2.7/build
creating build/temp.macosx-10.11-x86_64-2.7/Users
creating build/temp.macosx-10.11-x86_64-2.7/Users/zhangxin
creating build/temp.macosx-10.11-x86_64-2.7/Users/zhangxin/github
creating build/temp.macosx-10.11-x86_64-2.7/Users/zhangxin/github/warp-ctc
creating build/temp.macosx-10.11-x86_64-2.7/Users/zhangxin/github/warp-ctc/pytorch_binding
creating build/temp.macosx-10.11-x86_64-2.7/Users/zhangxin/github/warp-ctc/pytorch_binding/src
clang -fno-strict-aliasing -fno-common -dynamic -g -O2 -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -I/usr/local/lib/python2.7/site-packages/torch/utils/ffi/../../lib/include -I/usr/local/lib/python2.7/site-packages/torch/utils/ffi/../../lib/include/TH -I/Users/zhangxin/github/warp-ctc/include -I/usr/local/Cellar/python/2.7.13/Frameworks/Python.framework/Versions/2.7/include/python2.7 -c build/_warp_ctc.c -o build/temp.macosx-10.11-x86_64-2.7/build/_warp_ctc.o -std=c++11 -fPIC
error: invalid argument '-std=c++11' not allowed with 'C/ObjC'
error: command 'clang' failed with exit status 1