xiaofengshi / chinese-ocr Goto Github PK

在训练自己的模型的时候（keys修改过）发现得到的结果总是00或000。
是否预训练模型不是用trainbatch.py训练出来的？

我用的是keras的ocr代码，里面对样本的处理存在以下差异：

1. ocr/model.predict方法只是把图像resize为(32,width)，之后reshape为(32,width,1)，将数据处理为0，1之间。

2. 而trainbatch.py和dataset.py中会在resize后把数据reshape为(-1,32,256,1)。之后处理为0，1之间之后，再减0.5，再除0.5。

3. train.py中是Length = int(imgW/4)-1，trainbatch.py中是Length = int(imgW/4)-2

这样trainbatch.py，train.py和ocr/model.predict三个地方的对样本的处理方式不是都不同吗？好像只有allinonetrain.py与trainbatch.py一致。

是因为这个原因导致我用自己的keys.py训练的时候总是得不到想要的效果吗？

中文OCR的数据集怎么生成的? 定位没啥问题,就是中文ocr识别率比较低

我用自己的中文训练集训练的时候，会报这个错：

RuntimeError: CUDA error: an illegal memory access was encountered

然后我调试发现这个：

这个地方，应该是这样吧：

length = [len(s) for s in text]

不然把字符串拼起来拆分不开了吧？我改完就可以跑了

您在训练的时候样本数是多少， model.fit_generator(
gen(train_loader, flag='train'),
steps_per_epoch=102400,
epochs=200,
validation_data=gen(test_loader, flag='test'),
callbacks=[checkpointer, rlu],
validation_steps=1024)
题主这儿参数设置是不是有问题啊，steps_per_epoch不是等于 样本总数/batch 吗？

7f28bde41000-7f28bde48000 rw-p 00019000 08:05 6001167 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5s.cpython-36m-x86_64-linux-gnu.so
7f28bde48000-7f28bde68000 r-xp 00000000 08:05 6001151 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5a.cpython-36m-x86_64-linux-gnu.so
7f28bde68000-7f28be068000 ---p 00020000 08:05 6001151 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5a.cpython-36m-x86_64-linux-gnu.so
7f28be068000-7f28be06b000 rw-p 00020000 08:05 6001151 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5a.cpython-36m-x86_64-linux-gnu.so
7f28be06b000-7f28be06c000 rw-p 00000000 00:00 0
7f28be06c000-7f28be06f000 rw-p 00024000 08:05 6001151 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5a.cpython-36m-x86_64-linux-gnu.so
7f28be06f000-7f28be077000 r-xp 00000000 08:05 6001144 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5z.cpython-36m-x86_64-linux-gnu.so
7f28be077000-7f28be277000 ---p 00008000 08:05 6001144 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5z.cpython-36m-x86_64-linux-gnu.so
7f28be277000-7f28be27b000 rw-p 00008000 08:05 6001144 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/h5py/h5z.cpython-36m-x86_64-linux-gnu.so
7f28be27b000-7f28be36b000 r-xp 00000000 08:05 547530678 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libgfortran-ed201abd.so.3.0.0
7f28be36b000-7f28be56a000 ---p 000f0000 08:05 547530678 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libgfortran-ed201abd.so.3.0.0
7f28be56a000-7f28be56c000 rw-p 000ef000 08:05 547530678 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libgfortran-ed201abd.so.3.0.0
7f28be56c000-7f28be56d000 rw-p 00000000 00:00 0
7f28be56d000-7f28be575000 rw-p 000f2000 08:05 547530678 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libgfortran-ed201abd.so.3.0.0
7f28be575000-7f28c0952000 r-xp 00000000 08:05 547530677 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libopenblasp-r0-8dca6697.3.0.dev.so
7f28c0952000-7f28c0b51000 ---p 023dd000 08:05 547530677 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libopenblasp-r0-8dca6697.3.0.dev.so
7f28c0b51000-7f28c0b71000 rw-p 023dc000 08:05 547530677 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libopenblasp-r0-8dca6697.3.0.dev.so
7f28c0b71000-7f28c0bd4000 rw-p 00000000 00:00 0
7f28c0bd4000-7f28c0c73000 rw-p 02501000 08:05 547530677 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/.libs/libopenblasp-r0-8dca6697.3.0.dev.so
7f28c0c73000-7f28c0e4d000 r-xp 00000000 08:05 283713113 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/core/multiarray.cpython-36m-x86_64-linux-gnu.so
7f28c0e4d000-7f28c104c000 ---p 001da000 08:05 283713113 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/core/multiarray.cpython-36m-x86_64-linux-gnu.so
7f28c104c000-7f28c1066000 rw-p 001d9000 08:05 283713113 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/core/multiarray.cpython-36m-x86_64-linux-gnu.so
7f28c1066000-7f28c1084000 rw-p 00000000 00:00 0
7f28c1084000-7f28c108b000 rw-p 001f4000 08:05 283713113 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/numpy/core/multiarray.cpython-36m-x86_64-linux-gnu.so
7f28c108b000-7f28c1094000 r-xp 00000000 08:05 813485134 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/math.cpython-36m-x86_64-linux-gnu.so
7f28c1094000-7f28c1293000 ---p 00009000 08:05 813485134 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/math.cpython-36m-x86_64-linux-gnu.so
7f28c1293000-7f28c1295000 rw-p 00008000 08:05 813485134 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/math.cpython-36m-x86_64-linux-gnu.so
7f28c1295000-7f28c1355000 rw-p 00000000 00:00 0
7f28c1355000-7f28c1357000 r-xp 00000000 08:05 813485104 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/_heapq.cpython-36m-x86_64-linux-gnu.so
7f28c1357000-7f28c1557000 ---p 00002000 08:05 813485104 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/_heapq.cpython-36m-x86_64-linux-gnu.so
7f28c1557000-7f28c1559000 rw-p 00002000 08:05 813485104 /root/anaconda3/envs/chinese-ocr/lib/python3.6/lib-dynload/_heapq.cpython-36m-x86_64-linux-gnu.so
7f28c1559000-7f28c1599000 rw-p 00000000 00:00 0
7f28c1599000-7f28c1759000 r-xp 00000000 08:05 24840547 /lib/x86_64-linux-gnu/libc-2.23.so
7f28c1759000-7f28c1959000 ---p 001c0000 08:05 24840547 /lib/x86_64-linux-gnu/libc-2.23.so
7f28c1959000-7f28c195d000 r--p 001c0000 08:05 24840547 /lib/x86_64-linux-gnu/libc-2.23.so
7f28c195d000-7f28c195f000 rw-p 001c4000 08:05 24840547 /lib/x86_64-linux-gnu/libc-2.23.so
7f28c195f000-7f28c1963000 rw-p 00000000 00:00 0
7f28c1963000-7f28c1a6b000 r-xp 00000000 08:05 24478636 /lib/x86_64-linux-gnu/libm-2.23.so
7f28c1a6b000-7f28c1c6a000 ---p 00108000 08:05 24478636 /lib/x86_64-linux-gnu/libm-2.23.so
7f28c1c6a000-7f28c1c6b000 r--p 00107000 08:05 24478636 /lib/x86_64-linux-gnu/libm-2.23.so
7f28c1c6b000-7f28c1c6c000 rw-p 00108000 08:05 24478636 /lib/x86_64-linux-gnu/libm-2.23.so
7f28c1c6c000-7f28c1c73000 r-xp 00000000 08:05 24478661 /lib/x86_64-linux-gnu/librt-2.23.so
7f28c1c73000-7f28c1e72000 ---p 00007000 08:05 24478661 /lib/x86_64-linux-gnu/librt-2.23.so
7f28c1e72000-7f28c1e73000 r--p 00006000 08:05 24478661 /lib/x86_64-linux-gnu/librt-2.23.so
7f28c1e73000-7f28c1e74000 rw-p 00007000 08:05 24478661 /lib/x86_64-linux-gnu/librt-2.23.so
7f28c1e74000-7f28c1e76000 r-xp 00000000 08:05 24478667 /lib/x86_64-linux-gnu/libutil-2.23.so
7f28c1e76000-7f28c2075000 ---p 00002000 08:05 24478667 /lib/x86_64-linux-gnu/libutil-2.23.so
7f28c2075000-7f28c2076000 r--p 00001000 08:05 24478667 /lib/x86_64-linux-gnu/libutil-2.23.so
7f28c2076000-7f28c2077000 rw-p 00002000 08:05 24478667 /lib/x86_64-linux-gnu/libutil-2.23.so
7f28c2077000-7f28c207a000 r-xp 00000000 08:05 24478630 /lib/x86_64-linux-gnu/libdl-2.23.so
7f28c207a000-7f28c2279000 ---p 00003000 08:05 24478630 /lib/x86_64-linux-gnu/libdl-2.23.so
7f28c2279000-7f28c227a000 r--p 00002000 08:05 24478630 /lib/x86_64-linux-gnu/libdl-2.23.so
7f28c227a000-7f28c227b000 rw-p 00003000 08:05 24478630 /lib/x86_64-linux-gnu/libdl-2.23.so
7f28c227b000-7f28c2293000 r-xp 00000000 08:05 24840577 /lib/x86_64-linux-gnu/libpthread-2.23.so
7f28c2293000-7f28c2492000 ---p 00018000 08:05 24840577 /lib/x86_64-linux-gnu/libpthread-2.23.so
7f28c2492000-7f28c2493000 r--p 00017000 08:05 24840577 /lib/x86_64-linux-gnu/libpthread-2.23.so
7f28c2493000-7f28c2494000 rw-p 00018000 08:05 24840577 /lib/x86_64-linux-gnu/libpthread-2.23.so
7f28c2494000-7f28c2498000 rw-p 00000000 00:00 0
7f28c2498000-7f28c2706000 r-xp 00000000 08:05 50528899 /root/anaconda3/envs/chinese-ocr/lib/libpython3.6m.so.1.0
7f28c2706000-7f28c2906000 ---p 0026e000 08:05 50528899 /root/anaconda3/envs/chinese-ocr/lib/libpython3.6m.so.1.0
7f28c2906000-7f28c296f000 rw-p 0026e000 08:05 50528899 /root/anaconda3/envs/chinese-ocr/lib/libpython3.6m.so.1.0
7f28c296f000-7f28c29a0000 rw-p 00000000 00:00 0
7f28c29a0000-7f28c29c6000 r-xp 00000000 08:05 24478619 /lib/x86_64-linux-gnu/ld-2.23.so
7f28c29c8000-7f28c2a08000 rw-p 00000000 00:00 0
7f28c2a16000-7f28c2a2a000 r-xp 00000000 08:05 555894194 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/tensorflow/python/framework/fast_tensor_util.so
7f28c2a2a000-7f28c2a2b000 r--p 00013000 08:05 555894194 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/tensorflow/python/framework/fast_tensor_util.so
7f28c2a2b000-7f28c2a2d000 rw-p 00014000 08:05 555894194 /root/anaconda3/envs/chinese-ocr/lib/python3.6/site-packages/tensorflow/python/framework/fast_tensor_util.so
7f28c2a2d000-7f28c2bb2000 rw-p 00000000 00:00 0
7f28c2bc0000-7f28c2bc1000 rw-p 00000000 00:00 0
7f28c2bc1000-7f28c2bc2000 r--s 00000000 00:05 64199 /dev/nvidia1
7f28c2bc2000-7f28c2bc3000 r--s 00000000 00:05 64194 /dev/nvidia0
7f28c2bc3000-7f28c2bc4000 rwxp 00000000 00:00 0
7f28c2bc4000-7f28c2bc5000 rw-p 00000000 00:00 0
7f28c2bc5000-7f28c2bc6000 r--p 00025000 08:05 24478619 /lib/x86_64-linux-gnu/ld-2.23.so
7f28c2bc6000-7f28c2bc7000 rw-p 00026000 08:05 24478619 /lib/x86_64-linux-gnu/ld-2.23.so
7f28c2bc7000-7f28c2bc8000 rw-p 00000000 00:00 0
7fffc5153000-7fffc5182000 rw-p 00000000 00:00 0 [stack]
7fffc51e4000-7fffc51e6000 r-xp 00000000 00:00 0 [vdso]
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall]
Aborted

请问大家有遇到这样的问题吗？是如何解决的呢，感谢~
from torch._C import *
ImportError: dlopen: cannot load any more object with static TLS

hi @xiaofengShi
为什么crnn 显示了网络结构之后 ，一直卡住不动？请问你的用Python3能训练吗？

请问这个该如何解决。。？

D:\Anaconda3\lib\site-packages\h5py_init_.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type.
from ._conv import register_converters as _register_converters
Using TensorFlow backend.
Traceback (most recent call last):
File "demo.py", line 8, in
import model
File "D:\git\CHINESE-OCR\model.py", line 15, in
from ctpn.text_detect import text_detect
File "D:\git\CHINESE-OCR\ctpn\text_detect.py", line 4, in
from .ctpn.model import ctpn
File "D:\git\CHINESE-OCR\ctpn\ctpn\model.py", line 10, in
from lib.fast_rcnn.config import cfg
ModuleNotFoundError: No module named 'lib.fast_rcnn'

ctpn/lib/utils/bbox.py
感觉这整个文件都有问题，难道不是用Python？用的是cython?

作者你好，我使用textgen.py生成了170w训练集和17w测试级，字体大小为20-50像素，语料采用的搜狗语料，背景是全白，用了15种字体，每张图片都是10个字符。我训练出来的accuracy=0.99，然后我在使用./test/test.png这张图片去测试，大致上估算了下正确率只有76%，然后添加背景进行训练其他条件不变，得到model accuracy=0.81。实际去测试的时候效果还不如第一个。我觉得问题出来数据集上面，但是训练字体，和测试字体都是标准的打印字体。请问下有什么建议，或者应该如何优化这个模型呢？

我只测试ｄｅｍｏ就出现很多问题：ImportError: dlopen: cannot load any more object with static TLS

Dear @xiaofengShi ,
I try to run demo.py and an error output: No module named 'torch'. Because ./crnn/crnn.py need to import torch.
Is Pytorch a must to run demo.py? As to my understanding, either Keras or Pytorch is OK to finish this.
Look forward to you reply. Thanks!

文字方向检测
基于图像分类，在VGG16模型的基础上，训练0、90、180、270度检测的分类模型，详细代码参考angle/predict.py文件，训练图片8000张，准确率88.23%。 模型地址百度云下载

损失函数ctc-loss在哪里调用了？

还是原来TOTAL里的那个模型

Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/common_shapes.py", line 686, in _call_cpp_shape_fn_impl
input_tensors_as_shapes, status)
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/errors_impl.py", line 473, in exit
c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.InvalidArgumentError: Dimension 1 in both shapes must be equal, but are 11 and 5531. Shapes are [512,11] and [512,5531]. for 'Assign_36' (op: 'Assign') with input shapes: [512,11], [512,5531].

作者,您好!我训练自己的数据,10个数字加1一个背景类,然后就报这个错误,不知道怎么修改?求解作者.

请问对图片/test/IMG_1563.JPG，/test/008.jpg 或者其他字体的/test/006.jpg 可以识别吗？我用xiaofengShi提供的model 识别不出文字，出现乱码。请问是要自己训练吗？有人训练过效果如何呢？

下了checkpoints权重，但是'NoneType' object has no attribute 'model_checkpoint_path'

您好，请问那个角度检测的数据集您有吗8000

你好,请问我训练自己的数据,也需要下载预训练权重文件吗?

请问博主，classify文字方向分类器的数据集是什么呢？是ICDAR 2011这类数据集吗

为什么trainbatch的训练速度比train的训练速度快这么多？

编译器识别不出关键字cimport

crnn训练时，我用Keras，在CPU上可以运行，用GPU运行时，会报错。不知道有没有遇到同样问题的小伙伴，可以加qq 2541899447一起讨论。报错原因：ValueError: Shape must be rank 1 but is rank 4 for 'batch_normalization_1/cond/FusedBatchNorm' (op: 'FusedBatchNorm') with input shapes: [?,4,?,512], [1,4,1,1], [1,4,1,1], [1,4,1,1], [1,4,1,1]
我估计是batchnormlization出错。

setup就没成功...

pytorch训练过程中，会在utils.py里的encode()函数卡住，不停迭代
def encode(self, text, depth=0):
"""Support batch or single str."""
if isinstance(text, str):
text = [self.dict[char.lower()] for char in text]
length = [len(text)]
if isinstance(text, str):
text = [self.dict.get(char, 0) for char in text]
length = [len(text)]
######## add for unicode
# elif isinstance(text, unicode):
# text = [self.dict.get(char, self.dict[u'-']) for char in text]
# length = [len(text)]
elif isinstance(text, collections.Iterable):
length = [len(text)]
print(length) #试着打印这个length 会一直增加
text = ''.join(str(v) for v in text)
text, _ = self.encode(text)
if depth:
return text, len(text)
return (torch.IntTensor(text), torch.IntTensor(length))

这个函数是有什么问题吗？@xiaofengShi

modelPath = os.path.join(os.getcwd(), "ocr/ocr0.2.h5")，这里的ocr0.2.h5文件在哪里下载呢？

如果训练全体汉字，大概需要多少张图片，迭代多少次，选用sgd还是adadelta？大家可以一起讨论啊

你好，我看readme里只写了如何训练区域识别ctpn和文本识别crnn两个网络，
实际使用时发现文字方向检测的准确率反而不是很高，
经常有一张图片旋转180°后还是识别成旋转前角度的情况，
请问这个部分怎么继续训练，提高准确率？

训练时候运行crnn_main.py
Traceback (most recent call last):
File "crnn_main.py", line 310, in
cost = trainBatch(crnn, criterion, optimizer)
File "crnn_main.py", line 257, in trainBatch
data = train_iter.next()
File "/root/anaconda3/envs/xiaofengshi-train2.7/lib/python2.7/site-packages/torch/utils/data/dataloader.py", line 336, in next
return self._process_next_batch(batch)
File "/root/anaconda3/envs/xiaofengshi-train2.7/lib/python2.7/site-packages/torch/utils/data/dataloader.py", line 357, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
TypeError: function takes exactly 5 arguments (1 given)

自己定制了一下， 但是发现在加载模型的时候出现placeholder 不在图中。 将每个模型单独测试时没问题的，放在一起就报错，应该是tensorflow多个模型加载时计算图的冲突，试了一下创建多个计算图但是无果，请问大佬这个该如何解决?

6735
/usr/local/lib/python3.5/dist-packages/h5py/init.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type.
from ._conv import register_converters as _register_converters
2018-11-03 21:43:19.268938: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2018-11-03 21:43:19.378460: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-11-03 21:43:19.378777: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1344] Found device 0 with properties:
name: GeForce GTX 1080 major: 6 minor: 1 memoryClockRate(GHz): 1.8475
pciBusID: 0000:01:00.0
totalMemory: 7.92GiB freeMemory: 7.50GiB
2018-11-03 21:43:19.378810: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1423] Adding visible gpu devices: 0
2018-11-03 21:43:19.635578: I tensorflow/core/common_runtime/gpu/gpu_device.cc:911] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-11-03 21:43:19.635627: I tensorflow/core/common_runtime/gpu/gpu_device.cc:917] 0
2018-11-03 21:43:19.635637: I tensorflow/core/common_runtime/gpu/gpu_device.cc:930] 0: N
2018-11-03 21:43:19.635920: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1041] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 7241 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080, pci bus id: 0000:01:00.0, compute capability: 6.1)
Tensor("Placeholder:0", shape=(?, ?, ?, 3), dtype=float32)
Tensor("conv5_3/conv5_3:0", shape=(?, ?, ?, 512), dtype=float32)
Tensor("rpn_conv/3x3/rpn_conv/3x3:0", shape=(?, ?, ?, 512), dtype=float32)
WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/tensorflow/contrib/learn/python/learn/datasets/base.py:198: retry (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.
Instructions for updating:
Use the retry module or similar alternatives.
Tensor("lstm_o/Reshape_2:0", shape=(?, ?, ?, 512), dtype=float32)
Tensor("lstm_o/Reshape_2:0", shape=(?, ?, ?, 512), dtype=float32)
Tensor("rpn_cls_score/Reshape_1:0", shape=(?, ?, ?, 20), dtype=float32)
Tensor("rpn_cls_prob:0", shape=(?, ?, ?, ?), dtype=float32)
Tensor("Reshape_2:0", shape=(?, ?, ?, 20), dtype=float32)
Tensor("rpn_bbox_pred/Reshape_1:0", shape=(?, ?, ?, 40), dtype=float32)
Tensor("Placeholder_1:0", shape=(?, 3), dtype=float32)
Loading network VGGnet_test... Restoring from text_detection_ctpn/checkpoints/VGGnet_fast_rcnn_iter_50000.ckpt... done done.

start to recognize : imgs/001.jpg
2018-11-03 21:43:25.242278: E tensorflow/stream_executor/cuda/cuda_dnn.cc:396] Loaded runtime CuDNN library: 7102 (compatibility version 7100) but source was compiled with 7005 (compatibility version 7000). If using a binary install, upgrade your CuDNN library to match. If building from sources, make sure the library loaded at runtime matches a compatible version specified during compile configuration.
2018-11-03 21:43:25.243118: F tensorflow/core/kernels/conv_ops.cc:712] Check failed: stream->parent()->GetConvolveAlgorithms( conv_parameters.ShouldIncludeWinogradNonfusedAlgo(), &algorithms)
已放弃 (核心已转储)

我再使用预训练模型的时候会遇到一个错误，我修改了key.py我训练了10个汉字，训练的时候会报错，RuntimeError: While copying the parameter named rnn.1.embedding.weight, whose dimensions in the model are torch.Size([11, 512]) and whose dimensions in the checkpoint are torch.Size([5530, 512]).
请问我如何修改将输出层去掉，然后初始化我的输出层。

1.需要安装几个库
apt install graphviz
pip3 install graphviz
pip3 install pydot
pip3 install torchvision
pip3 install torch torchvision
2.在运行sh的时候需要注意是python还是python3
3.更改
（1）需要将CHINESE-OCR-master\angle\predict.py文件中
/Users/xiaofeng/Code/Github/dataset/CHINESE_OCR/angle/modelAngle.h5改成
/opt/CHINESE-OCR-master/angle/modelAngle.h5
（2）需要将CHINESE-OCR-master\ctpn\ctpn\model.py文件中
/Users/xiaofeng/Code/Github/dataset/CHINESE_OCR/ctpn/ctpn_checkpoints/ 改成
/opt/CHINESE-OCR-master/ctpn/checkpoints/
4.调试，具体问题可以发送到我的邮件：### [email protected] 进行沟通。

ctpn下的应该是少了一个lib文件夹

请问作者是用多种字体进行crnn模型训练吗？？

你好，我用你代码中的ctpn与https://github.com/eragonruan/text-detection-ctpn都试了下，可能是因为表单文字较小，会有10%左右的文字没有识别出来（包含汉字、数字、英文）。我试着改了下config中的scale，识别出来的区域会有变化，但仍然会有一部分没有识别出来。

想请问下，针对这种情况，我是需要重新针对表单文字进行训练，还是可以通过调节ctpn参数达到较好的效果？非常感谢！

这个模型的训练需要花费大量内存吗？

warp-ctc是不是只支持linux？不支持win10安装吗？

安装还是不成功，请问 要怎样修改？

生成项目“test_gpu.vcxproj”的操作 - 失败。 生成项目“test_cpu.vcxproj”的操作 - 失败。 这2个不影响吧？

warp-ctc是不是只支持linux？不支持win10安装吗？ 谢谢 谢谢

NotFoundError (see above for traceback): Key lstm_o/bidirectional_rnn/bw/lstm_cell/biases not found in checkpoint

您好，我在运行demo.py的时候出现的这个问题一直没有找到，请问有跑通的同胞们分享以下怎么解决这个问题呢？
File "/home/xshine6/Downloads/CHINESE-OCR/ctpn/ctpn/model.py", line 37, in load_tf_model reader = tf.train.NewCheckpointReader(ckpt.model_checkpoint_path) AttributeError: 'NoneType' object has no attribute 'model_checkpoint_path'

我这边使用modelAngle.h5里的参数后文本方向检测的效果并不是很好，经常方向调整错误，所以想问一下这个参数是怎么训练的？该准备怎样的训练数据和模型？能否告知一下，谢谢

你好,我想问一下训练CRNN自己的数据,图片的命名格式是怎么样的?