#coding: utf-8
import torch
from torch.autograd import Variable
import utils
import dataset
import os
from PIL import Image

import models.crnn as crnn

#os.environ["CUDA_VISIBLE_DEVICES"] ="1"
model_path = './data/netCRNN_ch_nc_21_nh_128.pth'
img_path = './data/image33.jpg'
alphabet = u''ACIMRey万下依口哺摄次状璐癌草血运重'
#print(alphabet)
nclass = len(alphabet) + 1

判断是否含有GPU

if torch.cuda.is_available():
model = crnn.CRNN(32, 1, nclass, 128).cuda()
pre_model = torch.load(model_path)
else:
model = crnn.CRNN(32, 1, nclass, 128)
pre_model = torch.load(model_path,map_location=lambda storage, loc: storage)

print('loading pretrained model from %s' % model_path)
for k,v in pre_model.items():
print(k,len(v))
model.load_state_dict(pre_model)

converter = utils.strLabelConverter(alphabet)

transformer = dataset.resizeNormalize((100, 32))
image = Image.open(img_path).convert('L')

#是否含有GPU
if torch.cuda.is_available():
image = transformer(image).cuda()
else:
image = transformer(image)

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)
print('%-20s => %-20s' % (raw_pred.encode('utf8'), sim_pred.encode('utf8')))

@YoungMiao ，我从另外一个 issue 看到你提供了百度云的中文语料库的下载地址：http://pan.baidu.com/s/1jHYJeh4 密码：fdtk。

我下载下来后，发现需要解压密码，请问解压密码是什么？谢谢！

感谢～

麻烦问一下，crnn, 我用多个gpu训练的loss感觉不对，与单gpu训练相同Loss值的模型，预测结果非常差，单gpu训练的模型预测是正确的。@ wulivicte，请问您遇到过这种问题吗，怎么解决的，非常感谢。

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 ?

in the code CTCLoss can't find these probabilites .
However l don't find where to print the output probabilities in CTCloss(). In __init__.py the CTC class is defined as follow :

class _CTC(Function):
    def forward(self, acts, labels, act_lens, label_lens):
        is_cuda = True if acts.is_cuda else False
        acts = acts.contiguous()
        loss_func = warp_ctc.gpu_ctc if is_cuda else warp_ctc.cpu_ctc
        grads = torch.zeros(acts.size()).type_as(acts)
        minibatch_size = acts.size(1)
        costs = torch.zeros(minibatch_size)
        loss_func(acts,
                  grads,
                  labels,
                  label_lens,
                  act_lens,
                  minibatch_size,
                  costs)
        self.grads = grads
        self.costs = torch.FloatTensor([costs.sum()])
        return self.costs

    def backward(self, grad_output):
        return self.grads, None, None, None


class CTCLoss(Module):
    def __init__(self):
        super(CTCLoss, self).__init__()

    def forward(self, acts, labels, act_lens, label_lens):
        """
        acts: Tensor of (seqLength x batch x outputDim) containing output from network
        labels: 1 dimensional Tensor containing all the targets of the batch in one sequence
        act_lens: Tensor of size (batch) containing size of each output sequence from the network
        act_lens: Tensor of (batch) containing label length of each example
        """
        _assert_no_grad(labels)
        _assert_no_grad(act_lens)
        _assert_no_grad(label_lens)
        return _CTC()(acts, labels, act_lens, label_lens)

Hello，This model is RARE?Why it is CRNN in here.

您好，我看textbox实验的内容，环境配置的问题，请教如何配置呢？Ubuntu14.04系统下配安装docker？“服务器上用nvidia-docker从镜像gds/keras-th-tf-opencv中新建了caffe_ys容器。按照caffe的依赖文件，并编译GPU版本。”这句话不理解╮(╯▽╰)╭

我在运行 crnn_main.py的时候，因为我的类和预训练模型不一样，所以需要修改，我修改了key.py里面的汉字，再crnn_main修改requires_grad但是没找到修改类的地方，想请假一下在哪里修改啊

我上传图片中的数字都被识别成了英文字母，请问该怎么办呢？

Hi,
can anyone please tell me how to train this model for transfer learning.
My model is working fine. But when I train it I am getting following error.

Traceback (most recent call last): File "crnn_main.py", line 191, in <module> train_iter = iter(train_loader) File "/usr/local/lib/python2.7/dist-packages/torch/utils/data/dataloader.py", line 303, in __iter__ return DataLoaderIter(self) File "/usr/local/lib/python2.7/dist-packages/torch/utils/data/dataloader.py", line 143, in __init__ self.sample_iter = iter(self.sampler) File "/home/pranay/crnn.pytorch/dataset.py", line 99, in __iter__ random_start = random.randint(0, len(self) - self.batch_size) File "/usr/lib/python2.7/random.py", line 242, in randint return self.randrange(a, b+1) File "/usr/lib/python2.7/random.py", line 218, in randrange raise ValueError, "empty range for randrange() (%d,%d, %d)" % (istart, istop, width) ValueError: empty range for randrange() (0,-33, -33) Exception AttributeError: "'DataLoaderIter' object has no attribute 'shutdown'" in <bound method DataLoaderIter.__del__ of <torch.utils.data.dataloader.DataLoaderIter object at 0x7ff08e6da050>> ignored

I am using the following line on terminal to run the code
python crnn_main.py --trainroot="Pranay/train_set/" --valroot="Pranay/validation_set" --alphabet='0123456789abcdefghijklmnopqrstuvwxyz !-%.'"'"',#&$\/[]:()?;'

I just want to test that training is possible before I train the complete model. Hence I have selected validation set with 10 images and train set with 30 images. I know it is not possible to train by using just 30 images. But still, I just want to see my training on CPU working. I think this error has something to do with the small size of training data.

When I try to run for more images (150-200), my computer gets hanged. Hence, I am trying for small subset.

I am training on ICDAR 2015 data.
I generated .mdb data by using input images path list as (This is just a sample. I hae more data in both lists)

• ['../test_images/word_11.png',
  '../test_images/word_12.png',
  '../test_images/word_13.png',
  '../test_images/word_14.png',]

and label path list as
['Genaxis Theatre',
'[06]',
'62-03',
'Carpark',
]

I have generated train.mdb and lock.mdb files both in folder train_set and validation_set. I have changed the alphabet to take into account new special characters as
--alphabet='0123456789abcdefghijklmnopqrstuvwxyz !-%.'"'"',#&$/[]:()?;'

Please, can you help me train my model? Any help is really appreciated.

Thanks

打扰您，请问您的crnn模型涵盖的中文有多少个字，是常用字吗