search

xwjabc / hed

A PyTorch reimplementation of Holistically-Nested Edge Detection
170 stars 42 forks source link

Good edge detector. I want to improve the thickness of edge line. #29

Open yeongkwoncho opened 3 years ago

yeongkwoncho commented 3 years ago

Hello. I have tested this code and have got good results. Thank you for sharing. However, as I apply this code to my TEM images and get edge lines, the output images have rather thick edge lines. So, I need to post-process those outputs to use in my actual job. I need your idea. How can I get thinner edge lines from this algorithm?

xwjabc commented 3 years ago

Hi @yeongkwoncho, thank you for your interest in this repo! After you generate the thick edge lines, you can apply a non-maximum supression (NMS) operation on the edges and obtain thinner ones. The NMS part is in the MATLAB evaluation code. To have a glimpse, you can attempt to evaluate a pre-trained HED model on BSDS dataset with the MATLAB code. You will find the evaluation procedure generates a nms folder which contains the thinner edges after NMS.