Picture quality prediction on KonIQ 10K (val)
0.788SRCCFeedback Model
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Feedback ModelBackbone=ResNet-18, trained_on=PaQ-2-PiQ, fine-tuning=false2019.12 | 0.788 | 0.808 | |
| RoIPool ModelBackbone=ResNet-18, trained_on=PaQ-2-PiQ, fine-tuning=false2019.12 | 0.776 | 0.794 | |
| Baseline ModelBackbone=ResNet-18, trained_on=PaQ-2-PiQ, fine-tuning=false2019.12 | 0.753 | 0.764 | |
| NIMA(ResNet 18)Backbone=ResNet-18, trained_on=PaQ-2-PiQ, fine-tuning=false2019.12 | 0.707 | 0.679 | |
| NIMA(MobileNet v2)Backbone=MobileNet v2, trained_on=PaQ-2-PiQ, fine-tuning=false2019.12 | 0.666 | 0.721 |