SharifAmit/RVGAN

Hello!

Closed this issue · 5 comments

victory3398 commented

1667311930327
I meet i problem. I don't understand you how to solve the data number is small,and how to get the number 4320 for STARE、15120 for CHASE-DB1、4200 for DRIVE .can you explain it? or send me your preprocess datasets. Thanks!!!

SharifAmit commented

Hi,

Please check the following two codes : strided_crop_DRIVE.py and convert_npz_DRIVE.py

We have similar codes for CHASE and STARE as well in this repository.

Basically, we use a strided crop operation with resolution 128x128 to get the 4310, 15120 and 4200 images for STARE, CHASE and DRIVE dataset.

victory3398 commented

1667354925216

Tranks! But I have the second questioin,the test dataset images is original shape and reshape it to 128x128,not crop it,only with stride = 3?

SharifAmit commented

For inference, please check the following code.

For testing we do not resize the image to 128x128. Rather, we take overlapping strided crop to create test crops, however the stride size is small s=3 to get finer segmentation results. The reason is after you have all the prediction test crops, you stitch and combine to get the original sized test image.

victory3398 commented

Perfectly! I will try to do this work ,Thanks!!!

SharifAmit commented

Thanks, closing the issue.

Please reopen the issue if you have more questions !