A metric for Perceptual Image-Error Assessment through Pairwise Preference (PieAPP at CVPR 2018).
A metric for Perceptual Image-Error Assessment through Pairwise Preference (PieAPP at CVPR 2018).
Test results of official trained model(download_PieAPPv0.1_PT_weights.sh) not same with paper
Size mismatch when loading the pretrained pytorch model.
Available training code
question about hole-filled image
PieAPPv0.1.exe ignores --gpu_id parameter
Train with Custom dataset?
torch cant use numpy??