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
1.49k stars 318 forks source link

Hello,EDVR Train and val dataset question #38

Open SeriouslyHAO opened 5 years ago

SeriouslyHAO commented 5 years ago

Hello,how to know datasets in the lmdb is train dataset or val dataset???May be by num?datasets before 240 is train dataset,after 240 is val dataset???thx

yinnhao commented 5 years ago

line 53 in REDS_dataset.py self.paths_GT = [ v for v in self.paths_GT if v.split('_')[1] not in ['00033', '00092', '00027', '00165'] ]

xinntao commented 5 years ago
  1. We have used a different partition for training and testing. More details can be found here.
  2. As @hahahaprince has pointed out, we excluded the evaluation clips in the dataset. https://github.com/xinntao/EDVR/blob/e1e6c140819930a4f21bdd43d1f209c562a473f1/codes/data/REDS_dataset.py#L53-L56
  3. Different from what @hahahaprince shows in the codes, we use the 000, 011, 015 and 020 clips from the original training set as the testing clips.
SeriouslyHAO commented 5 years ago

Thx for reply.