search

jiwoon-ahn / psa

Learning Pixel-level Semantic Affinity with Image-level Supervision for Weakly Supervised Semantic Segmentation, CVPR 2018
MIT License
380 stars 62 forks source link

question on generated segmentation masks #23

Open Seohyeong opened 4 years ago

Seohyeong commented 4 years ago

Thanks for sharing this work! I'm rather new to segmentation. I wondered how I can convert generated segmentation masks (which are .npy files located at cam folder) to .png images since deeplab segmentation network seems to take .png masks as segmentation labels. Could you refer to any source code or reference regarding this?

Seohyeong commented 4 years ago

To be more concrete, segmentation masks that I've made running your code gives me .npy format files and it seems like probability value is assigned to each of the location(pixel) in the image. I'm wondering if there's a threshold value you used to separate the background.