search

xuebinqin / U-2-Net

The code for our newly accepted paper in Pattern Recognition 2020: "U^2-Net: Going Deeper with Nested U-Structure for Salient Object Detection."
Apache License 2.0
8.52k stars 1.47k forks source link

Problems related to the training set #16

Closed ljc19940403 closed 4 years ago

ljc19940403 commented 4 years ago

First，appreciate your excellent work！In the training set, I need a little guidance. In BASNet, you detailed the composition of the training set，In this job, did you use the same training set during training？

xuebinqin commented 4 years ago

Yes, the training set (DUTS-TR) is the same except for the input size.

On May 12, 2020, at 9:50 PM, kobe notifications@github.com wrote:

First，appreciate your excellent work！In the training set, I need a little guidance. In BASNet, you detailed the composition of the training set，In this job, did you use the same training set during training？

— 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/16, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORPTRNJUPJV4XLP4UZLRRIKH7ANCNFSM4M7MD2EA.

ljc19940403 commented 4 years ago

Thank you for your reply！u-2 also has excellent results in icon detection, this is great，Can I continue to improve model performance through other training data？

xuebinqin commented 4 years ago

I think so. The current u2net model is trained with only DUTS-TR (for the fairness in comparing with other existing models). More dataset should improve the performance more or less.