focal_segmentation icon indicating copy to clipboard operation
focal_segmentation copied to clipboard

Published 20 hours ago • verified publisher icon doiken23

Is focalloss2d implementation correct?

Open stillwaterman opened this issue 6 years ago • 1 comments

Hello, I found you use pt = torch.exp(logpt) to calculate probability, but which sample's probability is represented by pt. I think pt represents the probability of a sample, not the whole dataset.

stillwaterman avatar Dec 21 '18 08:12 stillwaterman

Hello, I think I came up with similar problem, have you ever solved this issue? Hoping for your response.

ZeroCSIS avatar Aug 09 '20 15:08 ZeroCSIS