Image Quality Assessment on AGIQA 3K (test)
0.893SRCCMA-AGIQA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MA-AGIQASupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.893 | 0.927 | — | |
| LEAFSupervision Paradigm=Weak-supervised, MOS labeling ratio=30% MOS2026.01 | 0.868 | 0.914 | — | |
| MANIQASupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.861 | 0.911 | — | |
| Q-AlignSupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.852 | 0.881 | — | |
| HyperIQASupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.85 | 0.904 | — | |
| CLIP-IQA+Opinion-Unaware=false2024.03 | 0.844 | 0.894 | — | |
| CLIP-IQA+Supervision Paradigm=Weak-supervised2026.01 | 0.844 | 0.894 | — | |
| LEAFSupervision Paradigm=Weak-supervised, MOS labeling ratio=10% MOS2026.01 | 0.841 | 0.899 | — | |
| StairIQASupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.834 | 0.893 | — | |
| DBCNNSupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.826 | 0.89 | — | |
| MUSIQSupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.82 | 0.865 | — | |
| CONTRIQUEOpinion-Unaware=false2024.03 | 0.817 | 0.879 | — | |
| CONTRIQUESupervision Paradigm=Weak-supervised2026.01 | 0.817 | 0.879 | — | |
| Re-IQAOpinion-Unaware=false2024.03 | 0.811 | 0.874 | — | |
| Re-IQASupervision Paradigm=Weak-supervised2026.01 | 0.811 | 0.874 | — | |
| GRepQOpinion-Unaware=false2024.03 | 0.807 | 0.858 | — | |
| GRepQ-DSupervision Paradigm=Weak-supervised2026.01 | 0.807 | 0.858 | — | |
| ARNIQAOpinion-Unaware=false2024.03 | 0.803 | 0.881 | — | |
| ARNIQASupervision Paradigm=Weak-supervised2026.01 | 0.803 | 0.881 | — | |
| ELIQSupervision=Weak-supervised2026.02 | 0.789 | 0.881 | — | |
| IPCESupervision=Weak-supervised2026.02 | 0.77 | 0.872 | — | |
| AGQG +IPCEIncorporates AGQG module=true, Training dataset=AIGCIQA20232025.12 | 0.767 | 0.826 | 0.797 | |
| AGQG +CLIP-AGIQAIncorporates AGQG module=true, Training dataset=AIGCIQA20232025.12 | 0.757 | 0.791 | 0.774 | |
| IPCEIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.752 | 0.809 | 0.781 | |
| LEAFSupervision Paradigm=Label-free2026.01 | 0.749 | 0.811 | — | |
| StairRewardSupervision=Weak-supervised2026.02 | 0.747 | 0.852 | — | |
| FEWSHOT-Q-ALIGNTraining Set=KonIQ, #Training=2K (20%)2023.12 | 0.74 | 0.791 | — | |
| Q-ALIGNTraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.735 | 0.772 | — | |
| ImageRewardSupervision=Weak-supervised2026.02 | 0.729 | 0.786 | — | |
| Q-ALIGNTraining Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.723 | 0.786 | — | |
| ELIQSupervision=Label-free2026.02 | 0.717 | 0.73 | — | |
| LIQETraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.708 | 0.772 | — | |
| CLIP-IQA+Supervision=Weak-supervised2026.02 | 0.704 | 0.738 | — | |
| FEWSHOT-Q-ALIGNTraining Set=SPAQ, #Training=2.2K (20%)2023.12 | 0.702 | 0.772 | — | |
| MDFSOpinion-Unaware=true2024.03 | 0.697 | 0.723 | — | |
| PickScoreSupervision=Weak-supervised2026.02 | 0.697 | 0.763 | — | |
| CLIP-IQA+Training Set=KonIQ, #Training=7K (70%)2023.12 | 0.685 | 0.736 | — | |
| DUBMASupervision Paradigm=Label-free2026.01 | 0.684 | 0.701 | — | |
| LIQETraining Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.672 | 0.722 | — | |
| MDFSSupervision Paradigm=Label-free2026.01 | 0.672 | 0.676 | — | |
| AGQG +CLIP-IQAIncorporates AGQG module=true, Training dataset=AIGCIQA20232025.12 | 0.672 | 0.727 | 0.7 | |
| AGQG +TIERIncorporates AGQG module=true, Training dataset=AIGCIQA20232025.12 | 0.671 | 0.716 | 0.694 | |
| QualiCLIPOpinion-Unaware=true2024.03 | 0.667 | 0.735 | — | |
| QualiCLIPSupervision Paradigm=Label-free2026.01 | 0.667 | 0.735 | — | |
| CLIP-AGIQAIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.66 | 0.679 | 0.67 | |
| HyperIQAIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.657 | 0.692 | 0.674 | |
| NIMATraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.654 | 0.715 | — | |
| AMFF-NetIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.654 | 0.695 | 0.674 | |
| TIERIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.653 | 0.672 | 0.663 | |
| TReSIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.646 | 0.702 | 0.674 | |
| DBCNNTraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.641 | 0.73 | — | |
| HyperIQATraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.64 | 0.702 | — | |
| CLIP-IQAOpinion-Unaware=true2024.03 | 0.638 | 0.711 | — | |
| CLIP-IQASupervision Paradigm=Label-free2026.01 | 0.638 | 0.711 | — | |
| MUSIQIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.635 | 0.673 | 0.654 | |
| HPSv1Supervision=Weak-supervised2026.02 | 0.634 | 0.7 | — | |
| MUSIQTraining Set=KonIQ, #Training=7K (70%)2023.12 | 0.63 | 0.722 | — | |
| DB-CNNIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.627 | 0.688 | 0.657 | |
| ARNIQA-OUOpinion-Unaware=true2024.03 | 0.614 | 0.679 | — | |
| CLIP-IQAIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.614 | 0.665 | 0.64 | |
| GRepQ-OUOpinion-Unaware=true2024.03 | 0.613 | 0.734 | — | |
| GRepQ-ZSupervision Paradigm=Label-free2026.01 | 0.613 | 0.734 | — | |
| ILNIQESupervision Paradigm=Label-free2026.01 | 0.609 | 0.655 | — | |
| CLIPSupervision=Weak-supervised2026.02 | 0.597 | 0.683 | — | |
| PickScoreSupervision=Label-free2026.02 | 0.593 | 0.622 | — | |
| CL-MIOpinion-Unaware=true2024.03 | 0.591 | 0.665 | — | |
| ImageRewardSupervision=Label-free2026.02 | 0.579 | 0.607 | — | |
| CLIP-IQA+Training Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.577 | 0.614 | — | |
| ResNet50Incorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.576 | 0.612 | 0.594 | |
| Fang et al.Training Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.57 | 0.649 | — | |
| MUSIQTraining Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.564 | 0.675 | — | |
| HPSv1Supervision=Label-free2026.02 | 0.562 | 0.577 | — | |
| NIMATraining Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.534 | 0.63 | — | |
| IL-NIQEOpinion-Unaware=true2024.03 | 0.528 | 0.544 | — | |
| NIQESupervision Paradigm=Label-free2026.01 | 0.523 | 0.566 | — | |
| NIQEOpinion-Unaware=true2024.03 | 0.51 | 0.526 | — | |
| ViT-B/32Incorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.509 | 0.571 | 0.54 | |
| CLIP-IQA+Supervision=Label-free2026.02 | 0.501 | 0.524 | — | |
| Re-IQAIncorporates AGQG module=false, Training dataset=AIGCIQA20232025.12 | 0.473 | 0.352 | 0.412 | |
| BRISQUESupervision Paradigm=Supervised, MOS labeling ratio=100%2026.01 | 0.472 | 0.561 | — | |
| CONTRIQUE-OUOpinion-Unaware=true2024.03 | 0.465 | 0.489 | — | |
| DBCNNTraining Set=SPAQ, #Training=8.8K (80%)2023.12 | 0.459 | 0.518 | — | |
| CLIPSupervision=Label-free2026.02 | 0.428 | 0.45 | — | |
| Re-IQA-OUOpinion-Unaware=true2024.03 | 0.366 | 0.43 | — |