Open JasonFinisher123 opened 3 years ago
@JasonFinisher123 Hi,i have met the same problem.Have you sloved this problem about segmentation of small targets? Thanks for your reply
U^2-Net resize the input image to 320x320. You have to double check if the resizing reduce the target size to much. If yes, you may have to use a large input size e.g. 640x640. If not, you can try to train more iterations, because it usually takes more time to converge to the small targets. If neither works, you may think about change of the loss weights of different classes to balance the training process.
On Fri, Oct 30, 2020 at 4:28 AM JasonFinisher123 notifications@github.com wrote:
HI,Thanks for your answer. I want to know that whether U2net is suitable for segmentation of small targets. I have trained my datasets which is small objects segmentation,but U2net predicted all 0, Nothing. Do i need change something ?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NathanUA/U-2-Net/issues/86, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORMJJ7GV5WKEXO4UMHTSNKIOLANCNFSM4TEZ2IVQ .
-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage:https://webdocs.cs.ualberta.ca/~xuebin/
HI,Thanks for your answer. I want to know that whether U2net is suitable for segmentation of small targets. I have trained my datasets which is small objects segmentation,but U2net predicted all 0, Nothing. Do i need change something ?