The result of retraining

xuebinqin / BASNet

Code for CVPR 2019 paper. BASNet: Boundary-Aware Salient Object Detection

MIT License

1.37k stars 250 forks source link

The result of retraining #61

Closed xiaobao886 closed 3 years ago

xiaobao886 commented 3 years ago

I tried to perform offline data enhancement according to the training method in the paper, to double the dataset of DUTS by horizontally flipping it. Use iterative 400K weights for testing and evaluation. MAE=0.67, But 0.47 in the paper，maxF and relaxF higher than in the paper. Excuse me, why is this？

xuebinqin commented 3 years ago

Replied the similar question in https://github.com/xuebinqin/U-2-Net/issues/240.