DR-Net: A Novel Generative Adversarial Network for Single Image Deraining

Table 1

Layer parameters of the generative sub-network.

Layer Name	Parameters
Input	Rain image
Layer 1	Conv1. (7,7,64), stride=1, padding=3; Tanh
Layer 2	Conv2. (3,3,64), stride=1, padding=1; Tanh
Layer 3	Conv3. (3,3,64), stride=1, padding=1; Tanh
Layer 4	Conv4. (3,3,64), stride=1, padding=1; Tanh
Layer 5; Add layer	Conv5. (3,3,64), stride=1, padding=1; Tanh; Concatenate (layer 1, layer 5)
Layer 6	Conv6. (3,3,64), stride=1, padding=1; Tanh
Layer 21	Conv21. (5,5,64), stride=1, padding=2; Tanh
Layer 31	Conv31. (5,5,64), stride=1, padding=2; Tanh
Layer 41	Conv41. (5,5,64), stride=1, padding=2; Tanh
Layer 51; Add layer	Conv51. (5,5,64), stride=1, padding=2; Tanh; Concatenate (layer 21, layer 51)
Layer 61	Conv61. (5,5,64), stride=1, padding=2; Tanh
Add layer	Concatenate (layer 1, layer 6, layer 61)
Layer 7	Conv7. (3,3,32), stride=1, padding=1; Tanh
Layer 8	Conv8. (3,3,32), stride=1, padding=1; Tanh
Output	Rain-removal image