-
Traceback (most recent call last):
File "train.py", line 191, in
train_iter = iter(train_loader)
File "C:\Users\ling\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 819, …
-
@Holmeyoung
in https://github.com/Holmeyoung/crnn-pytorch/issues/17 you mentioned that your codes only support training with text length 26.
![0](https://user-images.githubusercontent.com/62248916…
ghost updated
4 years ago
-
我用crnn.pytorch训练数字,训练的时候loss很低了:
888--22---6--11--00--66---44-99---4--11--11--11--00--66--66---33- => 8261064941110663 , gt: 8261064941110663
99----66--6---5--00--33--88--88--22--11--22--33--88--…
-
CRNN(
(cnn): Sequential(
(conv0): Conv2d(1, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
(relu0): ReLU(inplace=True)
(pooling0): MaxPool2d(kernel_size=2, stride=2, padding=…
-
when i run this command:-
train.sh 2
on my git bash it produces following results:-
Namespace(batch_size=16, bits=8, checkpoint_iters=10000, clip=0.5, decoder_fuse_ …
-
-
Can it support different English and numeric fonts and if not then how can i train it to do so.
-
image = torch.FloatTensor(opt.batchSize, 3, opt.imgH, opt.imgH)
i don't know why setting two parameters of opt.imgH,?
-
我现在用的是DB+CRNN的流程做OCR 全是python版本的
python版本crnn trt的模型我试着测了一下耗时 发现还比ONNX慢 不清楚是不是C++和python的关系
python版本db trt我测试比onnx快了大概3倍 但是还是得200多ms(之前做过yolo检测 转trt大概20ms C++版本)
有点疑惑是python和C++会差那么多么(我也都是动态输入)
…
-
试了pytorch 0.2, 0.4版本 都是同样错误。看了下貌似是不能同时shuffle和sampler,请问这个情况如何解决呢
@meijieru @ZhangXinNan @Weihnachtswurst
Traceback (most recent call last):
File "crnn_main.py", line 69, in
collate_fn=…