yfzhou

我参考mmocr，写了一个简单版本的，转onnx的时候有些问题，有人能看看吗？谢谢 https://github.com/Bourne-M/SDMG_R

> @Bourne-M Could you elaborate on the issue encountered and the steps to reproduce? 好的。 1.sdmg2pb.py->torch2pb 函数中定义了5个输入，但是导出的onnx中只有4个。 2.参照mmocr的实现，代码里有些对tensor的slice操作，在转onnx的时候会引发 warning。

@AllentDan could you help me

> LayoutReader is trained with left-to-right and top-to-bottom order so it is recommended to follow a similar setting in the real application. I assume the difference you mentioned is probably...

> @280868756 你好，我也发现了这个问题，同样的文字就是上下空白多了些识别的很差，空白小的是别的蛮好的，请问你解决这个问题了吗 去掉空白

感谢回复，我的环境是这样的： certifi 2016.9.26 cycler 0.10.0 h5py 2.7.1 matplotlib 2.1.1 numpy 1.11.1 opencv-python 3.1.0.0 Pillow 5.0.0 pip 20.0.2 pygame 1.9.3 pyparsing 2.2.0 python-dateutil 2.6.1 pytz 2017.3 scipy 1.0.0 setuptools 36.4.0 six...

> I think the issue is with the shape. Upon checking the code `center` and `ij` comes from: > > ``` > contour,hier = cv2.findContours(mask.copy().astype('uint8'), > mode=cv2.RETR_CCOMP, > method=cv2.CHAIN_APPROX_SIMPLE) >...

may i asking how should i get pictures generated out?

i have got the problem,just modify contour,hier = ... to _,contour,hier = , then it will work. but i wondering why its always timeout while generating pictures

> I think at some point I switched to using CentOS 7 because some of the preprocessing also has a Linux dependency. If we want a more detailed steps on...