release 0.1

Author: Zeqiang-Lai

README.md

@@ -1,2 +1,38 @@
 # MAN
-PyTorch Implementation of Mixed Attention Network for Hyperspectral Image Denoising
+
+PyTorch Implementation of [Mixed Attention Network for Hyperspectral Image Denoising]()
+
+🌟 Hightlights
+
+- We propose a new adaptive skip connection which weights different features elementwise.
+- We propose a lightweight mixed attention block with spectrum attention and channel attention.
+
+## Usage
+
+- Use our model.
+
+```python
+import torch
+from mans import man
+
+net = man()
+x = torch.randn(4,1,31,64,64)
+y = net(x)
+```
+
+- Training and Testing with [HSIR](https://github.com/bit-isp/HSIR).
+
+```shell
+python -m hsirun.test -a mans.man -r ckpt/man_gaussian.pth -t icvl_512_30 icvl_512_50
+python -m hsirun.train -a mans.man 
+```
+
+## Citations
+
+```bibtex
+@misc{lai_man2022,
+    title={Mixed Attention Network for Hyperspectral Image Denoising}, 
+    author={Zeqiang Lai, Ying Fu},
+    year={2022},
+}
+```

mans.py

@@ -112,7 +112,6 @@ def _rnn_step(self, z, f, h):
         return h_
 
     def forward(self, inputs):
-        print(self.reverse)
         h = None
         Z = self.ffn_f(inputs).tanh()
         F = self.ffn_w(inputs).sigmoid()

test.py

@@ -0,0 +1,7 @@
+import torch
+from mans import man
+
+net = man()
+x = torch.randn(2,1,31,64,64)
+y = net(x)
+print(y.shape)