run create_patches_fp.py , masks file is good， but segmentation results of stitches file are mismatched.

[mengxfei]

Hello, thank you very much for your excellent work! I try to run create_patches_fp.py, process my data. But I found that compared to the mask image, the labeled borders of this file don't match the tumor location, as shown in the picture.

I tried to process the images inside the CAMELYON16 dataset and this didn't happen, I don't know what's wrong, I hope to get your answer!
Sincerely thanks and best regards！

May

[GeneGee]

You may need try to lower the value of parameter 'a_t', such as 10 or below

[mengxfei]

You may need try to lower the value of parameter 'a_t', such as 10 or below

I originally used openslide to convert the image to tif format, now I changed to pyvips to convert it, and get the tif format files, run the create_patches_fp.py code, and the mask and stitches can be matched. But the results in summary I got from running main.py with CAMELYON16 dataset (399 samples), something is not right, the precision is even one in some folds.