search

xinntao / EDVR

Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.
https://github.com/xinntao/BasicSR
1.48k stars 320 forks source link

Could you please share the code with the loss of Understanding Deformable Alignment in Video Super-Resolution in EDVR? #175

Open wenchen4321 opened 3 years ago

wenchen4321 commented 3 years ago

I have read the paper named Understanding Deformable Alignment in Video Super-Resolution. I am very interested in the Offset-fidelity Loss. Could you please share the code with the Offset-fidelity Loss in EDVR?

xinntao commented 3 years ago

Hi @wenchen4321
Thanks for your interests. We are sorting out our codes. We will release when it is ready.

csbhr commented 3 years ago

I am also very interested in the offset-fidelity loss presented in "Understanding Deformable Alignment in Video Super-Resolution". It is awesome to use optical flow to stabilize the training process.

Could you please release the codes of this offset-fidelity loss, or share more implementation details, such as: How to determine the hyperparameters (i.e. "lamda" and "t" in Eq4 and Eq5 in the paper) ? Whether the offsets of each level will be constrained? How to distinguish the corresponding direction of optical flow when constraining the two directions of offsets?

Look forward to your reply, thank you!