Image Quality Assessment on CSIQ
0.969SRCTOPIQ
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TOPIQTrained on=KADID-10k2026.02 | 0.969 | 0.963 | — | — | |
| CSFIQA2024.11 | 0.967 | 0.973 | — | — | |
| BiRQATrained on=KADID-10k2026.02 | 0.967 | 0.966 | — | — | |
| DeepJSD(TIP25)Feature Domain=Feature comparison2026.04 | 0.967 | 0.963 | — | — | |
| SFF2016.08 | 0.9627 | 0.9643 | 0.8288 | 0.0695 | |
| ARNIQAType=SSL + LR2023.10 | 0.962 | 0.973 | — | — | |
| ARNIQAopinion-unaware=false, zero-shot=true2024.03 | 0.961 | 0.969 | — | — | |
| AAT-BiRQATrained on=KADID-10k2026.02 | 0.959 | 0.96 | — | — | |
| GMSD2016.08 | 0.957 | 0.9541 | 0.8129 | 0.0786 | |
| DeepDC(TIP25)Feature Domain=Feature comparison2026.04 | 0.957 | 0.941 | — | — | |
| MDSI2016.08 | 0.9569 | 0.9531 | 0.813 | 0.0795 | |
| DBIQA(TPAMI25)Feature Domain=Feature comparison2026.04 | 0.956 | 0.961 | — | — | |
| QFM-IQM2024.11 | 0.954 | 0.965 | — | — | |
| Causal Disentanglement (Ours)Feature Domain=Causal Disentanglement2026.04 | 0.954 | 0.966 | — | — | |
| CDINet2024.11 | 0.952 | 0.96 | — | — | |
| DeepCausal(CVPR25)Feature Domain=Feature comparison2026.04 | 0.952 | 0.949 | — | — | |
| AHIQTrained on=KADID-10k2026.02 | 0.951 | 0.955 | — | — | |
| ADISTSFeature Domain=Feature comparison2026.04 | 0.95 | 0.957 | — | — | |
| DeepWSD(MM22)Feature Domain=Feature comparison2026.04 | 0.95 | 0.941 | — | — | |
| Re-IQAType=SSL + LR2023.10 | 0.947 | 0.96 | — | — | |
| Re-IQA2024.11 | 0.947 | 0.96 | — | — | |
| MAD2016.08 | 0.9467 | 0.95 | 0.797 | 0.082 | |
| DB-CNNType=Supervised learning2023.10 | 0.946 | 0.959 | — | — | |
| CONTRIQUEopinion-unaware=false, zero-shot=true2024.03 | 0.946 | 0.96 | — | — | |
| Re-IQAopinion-unaware=false, zero-shot=true2024.03 | 0.946 | 0.954 | — | — | |
| DBCNN2024.11 | 0.946 | 0.959 | — | — | |
| DEIQT2024.11 | 0.946 | 0.963 | — | — | |
| SCQI2016.08 | 0.9434 | 0.9268 | 0.787 | 0.0986 | |
| DACNN2024.11 | 0.943 | 0.957 | — | — | |
| VSI2016.08 | 0.9423 | 0.9279 | 0.7857 | 0.0979 | |
| ADD-GSIM2016.08 | 0.9422 | 0.9342 | 0.7894 | 0.0937 | |
| CONTRIQUEType=SSL + LR2023.10 | 0.942 | 0.955 | — | — | |
| DSCSI2016.08 | 0.9417 | 0.9313 | 0.7787 | 0.0956 | |
| GMSDFeature Domain=Traditional2026.04 | 0.939 | 0.938 | — | — | |
| Su et al.Type=Supervised learning2023.10 | 0.935 | 0.952 | — | — | |
| SR-SIM2016.08 | 0.9319 | 0.925 | 0.7725 | 0.0997 | |
| FSIMc2016.08 | 0.931 | 0.9192 | 0.769 | 0.1034 | |
| Causal Disentanglement (Ours w/o f)Feature Domain=Causal Disentanglement2026.04 | 0.931 | 0.946 | — | — | |
| DISTSTrained on=KADID-10k2026.02 | 0.929 | 0.928 | — | — | |
| VSIFeature Domain=Traditional2026.04 | 0.929 | 0.912 | — | — | |
| LPIPS-AlexTraining Set=KADID-10K2026.02 | 0.928 | 0.893 | — | — | |
| NLPDFeature Domain=Traditional2026.04 | 0.926 | 0.913 | — | — | |
| PDLFeature Domain=Feature comparison2026.04 | 0.926 | 0.919 | — | — | |
| DSDFeature Domain=Feature comparison2026.04 | 0.925 | 0.925 | — | — | |
| HyperIQAType=Supervised learning2023.10 | 0.923 | 0.942 | — | — | |
| HyperIQA2024.11 | 0.923 | 0.942 | — | — | |
| TRESType=Supervised learning2023.10 | 0.922 | 0.942 | — | — | |
| TReS2024.11 | 0.922 | 0.942 | — | — | |
| IWSSIM2016.08 | 0.9213 | 0.9144 | 0.7529 | 0.1063 | |
| DISTS(TPAMI20)Feature Domain=Feature comparison2026.04 | 0.92 | 0.919 | — | — | |
| VIF2016.08 | 0.9195 | 0.9277 | 0.7537 | 0.098 | |
| MSSSIM2016.08 | 0.9133 | 0.8991 | 0.7393 | 0.1149 | |
| BiRQATrained on=PIPAL2026.02 | 0.912 | 0.913 | — | — | |
| MQAF-RTraining Set=KADID-10K2026.02 | 0.91 | 0.903 | — | — | |
| WaDIQaM-FRTrained on=KADID-10k2026.02 | 0.909 | 0.901 | — | — | |
| AAT-BiRQATrained on=PIPAL2026.02 | 0.909 | 0.909 | — | — | |
| TOPIQTrained on=PIPAL2026.02 | 0.908 | 0.908 | — | — | |
| MetaIQA2024.11 | 0.899 | 0.908 | — | — | |
| P2P-BM2024.11 | 0.899 | 0.902 | — | — | |
| VIFFeature Domain=Traditional2026.04 | 0.899 | 0.899 | — | — | |
| PieAPPTraining Set=KADID-10K2026.02 | 0.897 | 0.89 | — | — | |
| TOPIQ-FR(TIP24)Feature Domain=Feature comparison2026.04 | 0.894 | 0.894 | — | — | |
| PieAPPTrained on=KADID-10k2026.02 | 0.892 | 0.877 | — | — | |
| PieAPP(CVPR18)Feature Domain=Feature comparison2026.04 | 0.883 | 0.864 | — | — | |
| MS-SSIMFeature Domain=Traditional2026.04 | 0.879 | 0.875 | — | — | |
| LPIPS-VGGTrained on=KADID-10k2026.02 | 0.876 | 0.896 | — | — | |
| LPIPS(CVPR18)Feature Domain=Feature comparison2026.04 | 0.876 | 0.894 | — | — | |
| MUSIQ2024.11 | 0.871 | 0.893 | — | — | |
| AHIQTrained on=PIPAL2026.02 | 0.865 | 0.861 | — | — | |
| CLIP-IQA+opinion-unaware=false, zero-shot=true2024.03 | 0.862 | 0.882 | — | — | |
| DISTSTrained on=PIPAL2026.02 | 0.859 | 0.862 | — | — | |
| LPIPS-VGGTrained on=PIPAL2026.02 | 0.858 | 0.857 | — | — | |
| DeQA-ScoreTraining Dataset=KonIQ, SPAQ, KADID2026.01 | 0.857 | 0.9 | — | — | |
| MEON2024.11 | 0.852 | 0.864 | — | — | |
| IL-NIQEopinion-unaware=true, zero-shot=true2024.03 | 0.848 | 0.883 | — | — | |
| Q-DeepSight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 0.847 | 0.893 | — | — | |
| GRepQopinion-unaware=false, zero-shot=true2024.03 | 0.844 | 0.871 | — | — | |
| SSIMFeature Domain=Traditional2026.04 | 0.833 | 0.81 | — | — | |
| ARNIQA-OUopinion-unaware=true, zero-shot=true2024.03 | 0.832 | 0.81 | — | — | |
| Q-insightTraining Dataset=KonIQ, KADIS2026.01 | 0.824 | 0.87 | — | — | |
| Q-InsightCategory=MLLM-based, Training Dataset=KonIQ2025.10 | 0.824 | 0.87 | — | — | |
| Q-Insight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 0.824 | 0.87 | — | — | |
| GLIANetTraining Dataset=LIVE2026.05 | 0.823 | — | — | — | |
| ILNIQE2024.11 | 0.822 | 0.865 | — | — | |
| WaDIQaM-FRTrained on=PIPAL2026.02 | 0.822 | 0.834 | — | — | |
| BIECON2024.11 | 0.815 | 0.823 | — | — | |
| BRISQUE2024.11 | 0.812 | 0.748 | — | — | |
| LIQECategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.808 | 0.782 | — | — | |
| DB-CNNTraining Dataset=TID20132019.07 | 0.807 | — | — | — | |
| DeQA-ScoreTraining Dataset=KonIQ, SPAQ, KADID, PIPAL2026.01 | 0.807 | 0.863 | — | — | |
| PSNRFeature Domain=Traditional2026.04 | 0.807 | 0.792 | — | — | |
| Q-HawkeyeTraining Dataset=KonIQ2026.01 | 0.806 | 0.845 | — | — | |
| Q-HawkeyeMethodology=MLLM-Based Methods2026.01 | 0.806 | 0.845 | — | — | |
| PSNRTraining Set=KADID-10K2026.02 | 0.806 | 0.796 | — | — | |
| Q-DeepSightCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.806 | 0.843 | — | — | |
| QualiCLIPopinion-unaware=true, zero-shot=true2024.03 | 0.804 | 0.842 | — | — | |
| MDFSopinion-unaware=true, zero-shot=true2024.03 | 0.797 | 0.829 | — | — | |
| VisualQuality-R1Training Dataset=KADID, SPAQ2026.01 | 0.797 | 0.855 | — | — | |
| SHDIQATraining Dataset=LIVE2026.05 | 0.789 | — | — | — | |
| RALICategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.788 | 0.828 | — | — |