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.
in EDVR_arch.py and SRResNet_arch.py, align_corners is set to be False, while in TOF_arch.py, align_corners is set to be True, is there any reason for this?
For TOFlow, we tested the converted model with both settings and found that aligned_corners=True gives a better result. We therefore use aligned_corners=True for TOFlow.
in EDVR_arch.py and SRResNet_arch.py, align_corners is set to be False, while in TOF_arch.py, align_corners is set to be True, is there any reason for this?