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
8.31k stars 1.43k forks source link

自己数据训练的u2net,归一化尺寸320,但是抠图效果边缘锯齿明显 #383

Closed xiangfeidecao closed 4 weeks ago

xiangfeidecao commented 1 month ago
    # transform=transforms.Compose([
        # RescaleT(320),
        # RandomCrop(288),
        # ToTensorLab(flag=0)]))

自己数据训练的u2net,归一化尺寸320,但是抠图效果边缘锯齿明显。如果用640训练锯齿有很大改善。但是640的速度太慢,而且rembg这个公开库也是用的320输入,但是他们就没那么锯齿。不知道怎么回事。