search

jiwoon-ahn / irn

Weakly Supervised Learning of Instance Segmentation with Inter-pixel Relations, CVPR 2019 (Oral)
MIT License
519 stars 100 forks source link

Time cost of generating one pseudo instance mask #41

Open fcc315 opened 2 years ago

fcc315 commented 2 years ago

Hi,

After testing the IRNet, I found it takes about 3 seconds to generate one pseudo instance mask on my machine. I searched around and found no one mentioned the efficiency here, or even in the WSIS community. Or maybe I missed some paper/post.

I understand for the final goal the inference time matters, not the time of generating one pseudo instance mask. But is there any way that I can make it faster? Why people don't care about this?

Thanks