Xiefan Guo

Maybe you can modify the code as follows, it works for me. ``` model.netEN.module.load_state_dict(torch.load("EN.pth")['net']) model.netDE.module.load_state_dict(torch.load("DE.pth")['net']) model.netMEDFE.module.load_state_dict(torch.load("MEDFE.pth")['net']) ```

你好， 这里我们参考了[Zeng *et al.*, 2019] https://github.com/researchmm/PEN-Net-for-Inpainting. 祝工作顺利！ javey-q ***@***.***> 于2022年5月9日周一 16:46写道： > 作者你好，我在阅读您的代码的CFA模块RAL部分中有些地方没有看懂 > 比如代码中‘rate’的设置，以及‘background_kernel_size’设置为2*rate，而且这部分在论文中也没有提到 > 想问下这部分在哪有详细的叙述或者是参考了哪篇论文的实现 > > — > Reply to this email directly, view it on GitHub...

Hi, we follow the experimental settings of [EdgeConnect](https://arxiv.org/pdf/1901.00212.pdf). For CelebA, we cropped the center 178x178 of the images, then resized them to 256x256 using bilinear interpolation.

Hi, maybe it is caused by the preprocessing of the image. We follow the experimental settings of [EdgeConnect](https://arxiv.org/pdf/1901.00212.pdf). For CelebA, we cropped the center 178x178 of the images, then resized...