Haoyu Chen

delete folder 'bin' in dataset directory and regenerate .pt files

我的代码是基于老版本的 [EDSR](https://github.com/sanghyun-son/EDSR-PyTorch) 的，因此可以很轻松的转移到最新的 EDSR 的代码中（支持1.0版本以上的Pytorch），你可以使用最新的 [EDSR](https://github.com/sanghyun-son/EDSR-PyTorch) 的代码，然后使用我的模型文件（code/model/aan.py）即可。

`x = attention * ax[:,0].view(a,1,1,1) + non_attention * ax[:,1].view(a,1,1,1)` 您好，可以通过直接看 [ax](https://github.com/haoyuc/A2N/blob/210b38cbe721f0283aa5ad20c98373c5822f917a/code/model/aan.py#L106) 的值得到

关于注意力机制的问题，文中描述是这样的： > channel attention generates a 1D (R^{C×1×1}) channel-wise attention vector, > spatial attention generates a 2D (R^{1×H×W}) attention mask, > channel-spatial attention generates 3D (R^{C×H×W}) attention map spatial attention...

There is no public code for diffusion index at present. You can contact the author of LAM for more information, and here is the [project page](https://x-lowlevel-vision.github.io/lam.html) of LAM.

For the heatmap, we first average a feature map in the channel dimension and then visualize it. For example, ```python feature = torch.mean(feature,1)[0,:,:] ``` The "0" in [0,:,:] means we...

As mentioned in the paper: > The white area in the feature map indicates zero values, the red area indicates positive values, and the blue area indicates negative values. You...

Because the default setting of `plt.matshow` is green, you can change the color by setting `cmap=`, like `plt.matshow(feature, cmap='seismic')` @nan-rock

> Please u provide the versions of cuda, cuddn, torch, and torchvision. Thanks torch==0.4.1 torchvision==0.2.1 As for cuda and cuddn, since the ECS I used has been deleted, I'm not...

In fact, you can simply use the training code from [EDSR](https://github.com/sanghyun-son/EDSR-PyTorch), it supports PyTorch > 1.0, you only need to add our model code `A2N/code/model/aan.py` to EDSR `EDSR-PyTorch/src/model/` folder