Jing Leung

我们的论文给出了BVMR在LVW和CLWD上的各项量化指标；BVMR模型可以裁剪 SplitNet网络，训练获得

去掉SplitNet Refinement, 去掉Coarse Stage里的channel attention，剩下的细节比对这BVMR 论文

你试试别的pytorch版本看看

请参考scripts/train.sh 里面batch_size的参数

> For a task without mask (just has target, input(before watermark, after watermark)) How to use this code? > > As "ground-truth watermark mask M", in your paper formula(1) is...

> > > For a task without mask (just has target, input(before watermark, after watermark)) How to use this code? > > > As "ground-truth watermark mask M", in your...

> It seems that the performance is good. To further fine tune, you can improve the network or add more data to make your model robust.

Hi, we use nn.Sigmoid at 1. https://github.com/bcmi/SLBR-Visible-Watermark-Removal/blob/47c665f1855ab6624cd52b28cefa797a9c8b96f7/src/networks/blocks.py#L286 2. https://github.com/bcmi/SLBR-Visible-Watermark-Removal/blob/47c665f1855ab6624cd52b28cefa797a9c8b96f7/src/networks/blocks.py#L290

Thanks for your attention! We provide an alternative link in the ```Dataset``` section of https://github.com/bcmi/Awesome-Visible-Watermark-Removal.

We use 1 Titan RTX in the experiments, and it takes about 4-5 days to complete the training