Hao Feng

@an1018 Note that In the [DocUNet Benchmark](https://www3.cs.stonybrook.edu/~cvl/docunet.html), the '64_1.png' and '64_2.png' distorted images are rotated by 180 degrees, which do not match the GT documents. It is ignored by most...

@an1018 For your Q2, this performance is based on GeoTr. 

@an1018 For Q3 and Q4, to reproduce the above performance, please use the geometric rectified images rather than the illumination corrected images.

  This is version information for your reference. Besides, what is your performance based on Setting 2?

Hi, this is the [link](https://digi.bib.uni-mannheim.de/tesseract/) for Windows. Our enev is Windows. Hope to get your reply. This is the [link](https://tesseract-ocr.github.io/tessdoc/Installation.html) for Ubuntu, but we do not have a try.

@an1018 Thanks for your reply. For OCR evaluation, I think that you can compare the performance with the same environment, whether it is windows or ubuntu.

We print(shdmap.max(), shdmap.min()) using your code, and the results are like that:  How should we scale the values? @Enuvesta Hi. Did you reproduce good results？ > Do something like...

@sagniklp Thanks for your reply. For training the refinement network, the shape of the input image is 256*256, and the output of the shading map is 128*128. For inference, is...

Hi~, I am also very pretty eager to have this dataset downloaded, when could the server be OK for us to download the dataset Thank you~! @sagniklp

Thanks for your attention to our DocScanner. You need to prepare your distorted images and corresponding warping flow for rectification. Hope this help.