Image Quality Assessment on KonIQ-10k
0.953SRCCDeQA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DeQAModel Category=Specialized, Domain=In domain2025.12 | 0.953 | 0.941 | |
| DeQACategory=Specialized, Evaluation Domain=In-domain, Evaluation Protocol=Single-task2026.03 | 0.953 | 0.941 | |
| Q-Align*Model Category=Specialized, Domain=In domain, Retrained=true2025.12 | 0.941 | 0.94 | |
| Q-AlignCategory=Specialized, Evaluation Domain=In-domain, Source=Cited from UniPercept2026.03 | 0.941 | 0.94 | |
| TATARCategory=Unified2026.03 | 0.941 | 0.942 | |
| UniPerceptModel Category=Ours2025.12 | 0.94 | 0.949 | |
| MA-AGIQAType=data-driven, Supervision Setting=Supervised2026.02 | 0.933 | 0.948 | |
| Q-InsightModel Category=Specialized, Domain=In domain2025.12 | 0.933 | 0.916 | |
| Q-InsightCategory=Specialized, Evaluation Domain=In-domain, Evaluation Protocol=Single-task2026.03 | 0.933 | 0.916 | |
| MANIQAType=data-driven, Supervision Setting=Supervised2026.02 | 0.93 | 0.946 | |
| Q-AlignType=data-driven, Supervision Setting=Supervised2026.02 | 0.922 | 0.911 | |
| StairIQAType=data-driven, Supervision Setting=Supervised2026.02 | 0.92 | 0.936 | |
| Re-IQAType=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.914 | 0.923 | |
| ELIQType=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.912 | 0.924 | |
| HyperIQAType=data-driven, Supervision Setting=Supervised2026.02 | 0.904 | 0.915 | |
| UniPerceptCategory=Unified, Training Protocol=Retrained under our protocol2026.03 | 0.901 | 0.922 | |
| LEAFSupervision type=Weak-supervised, MOS percentage=30%2026.01 | 0.899 | 0.916 | |
| CLIP-IQA+Type=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.895 | 0.909 | |
| CLIP-IQA+Supervision type=Weak-supervised2026.01 | 0.895 | 0.909 | |
| CONTRIQUEType=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.894 | 0.906 | |
| CONTRIQUESupervision type=Weak-supervised2026.01 | 0.894 | 0.906 | |
| QualiCLIP+Opinion-Unaware=false2024.03 | 0.889 | 0.898 | |
| Re-IQAOpinion-Unaware=false2024.03 | 0.883 | 0.887 | |
| GRepQOpinion-Unaware=false2024.03 | 0.882 | 0.883 | |
| CONTRIQUEOpinion-Unaware=false2024.03 | 0.874 | 0.882 | |
| CLIP-IQA+Opinion-Unaware=false2024.03 | 0.873 | 0.89 | |
| ARNIQAOpinion-Unaware=false2024.03 | 0.869 | 0.883 | |
| ARNIQAType=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.869 | 0.883 | |
| ARNIQASupervision type=Weak-supervised2026.01 | 0.869 | 0.883 | |
| LEAFSupervision type=Weak-supervised, MOS percentage=10%2026.01 | 0.867 | 0.903 | |
| GRepQ-DType=data-driven, Supervision Setting=Weak-supervised2026.02 | 0.855 | 0.868 | |
| GRepQ-DSupervision type=Weak-supervised2026.01 | 0.855 | 0.868 | |
| DBCNNType=data-driven, Supervision Setting=Supervised2026.02 | 0.844 | 0.862 | |
| MG-IQATraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.83 | 0.856 | |
| MUSIQType=data-driven, Supervision Setting=Supervised2026.02 | 0.824 | 0.937 | |
| ELIQType=data-driven, Supervision Setting=Label-free2026.02 | 0.818 | 0.831 | |
| QualiCLIPOU (Opinion-Unaware)=true, Train Dataset=PIPAL, Evaluation Setting=Cross-dataset2024.03 | 0.817 | 0.838 | |
| QualiCLIPOpinion-Unaware=true2024.03 | 0.817 | 0.838 | |
| QualiCLIPOpinion-Unaware=true2024.03 | 0.817 | 0.838 | |
| QualiCLIPType=data-driven, Supervision Setting=Label-free2026.02 | 0.817 | 0.838 | |
| LIQEOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.813 | 0.801 | |
| VQ-R1Training Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.812 | 0.838 | |
| GRepQOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.81 | 0.818 | |
| LIQEOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.81 | 0.787 | |
| DBCNNOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.804 | 0.823 | |
| Q-InsightTraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.804 | 0.829 | |
| ARNIQAOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.798 | 0.81 | |
| Re-IQAOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.796 | 0.827 | |
| QwenVL-3-Instruct-32BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.796 | 0.838 | |
| QwenVL-3-32BCategory=Open-Source, Evaluation Domain=Cross-domain2026.03 | 0.796 | 0.838 | |
| DeQA-ScoreTraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.795 | 0.82 | |
| Q-AlignTraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.789 | 0.813 | |
| CLIP-IQA+Opinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.783 | 0.798 | |
| CONTRIQUEOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.777 | 0.777 | |
| LEAFSupervision type=Label-free2026.01 | 0.777 | 0.801 | |
| LIQETraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.776 | 0.802 | |
| CLIP-IQA+Opinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.769 | 0.808 | |
| TOPIQTraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.769 | 0.795 | |
| GRepQ-ZType=data-driven, Supervision Setting=Label-free2026.02 | 0.768 | 0.784 | |
| GRepQ-ZSupervision type=Label-free2026.01 | 0.768 | 0.784 | |
| QCNOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.762 | 0.835 | |
| QwenVL-2.5-Instruct-72BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.762 | 0.82 | |
| QwenVL-2.5-72BCategory=Open-Source, Evaluation Domain=Cross-domain2026.03 | 0.762 | 0.82 | |
| QwenVL-3-Instruct-8BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.761 | 0.822 | |
| TRESOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.759 | 0.758 | |
| MANIQATraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.758 | 0.783 | |
| Re-IQATraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.754 | 0.779 | |
| ARNIQASupervision type=Label-free2026.01 | 0.746 | 0.762 | |
| MUSIQTraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.746 | 0.775 | |
| ARNIQAOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.745 | 0.775 | |
| GRepQOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.742 | 0.75 | |
| HyperIQAOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.738 | 0.735 | |
| CONTRIQUEOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.736 | 0.749 | |
| HyperIQAOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.734 | 0.773 | |
| MDFSType=data-driven, Supervision Setting=Label-free2026.02 | 0.733 | 0.737 | |
| MDFSSupervision type=Label-free2026.01 | 0.733 | 0.737 | |
| Q-Insight*Model Category=Specialized, Domain=In domain, Retrained=true2025.12 | 0.733 | 0.75 | |
| Q-InsightCategory=Specialized, Evaluation Domain=In-domain, Source=Cited from UniPercept2026.03 | 0.733 | 0.75 | |
| UNIQUETraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.73 | 0.763 | |
| QCNOpinion-Unaware=false, Training Dataset=FLIVE2024.03 | 0.728 | 0.789 | |
| TRESOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.727 | 0.756 | |
| GLM-4.5-V-106BA12BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.721 | 0.765 | |
| GLM-4.5Category=Open-Source, Evaluation Domain=Cross-domain2026.03 | 0.721 | 0.765 | |
| DBCNNOU (Opinion-Unaware)=false, Train Dataset=PIPAL, Evaluation Setting=Cross-dataset2024.03 | 0.718 | 0.744 | |
| CLIP-IQATraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.712 | 0.741 | |
| QwenVL-2.5-Instruct-7BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.708 | 0.762 | |
| BRISQUEType=handcraft, Supervision Setting=Supervised2026.02 | 0.705 | 0.707 | |
| DBCNNOpinion-Unaware=false, training_dataset=CLIVE2024.03 | 0.703 | 0.755 | |
| DUBMAType=data-driven, Supervision Setting=Label-free2026.02 | 0.703 | 0.74 | |
| DUBMASupervision type=Label-free2026.01 | 0.703 | 0.74 | |
| CLIP-IQAType=data-driven, Supervision Setting=Label-free2026.02 | 0.695 | 0.727 | |
| CLIP-IQASupervision type=Label-free2026.01 | 0.695 | 0.727 | |
| GPT-4oModel Category=Proprietary, Domain=Cross domain testing2025.12 | 0.695 | 0.744 | |
| GPT-4oCategory=Proprietary, Evaluation Domain=Cross-domain2026.03 | 0.695 | 0.744 | |
| LIQEOU (Opinion-Unaware)=false, Train Dataset=PIPAL, Evaluation Setting=Cross-dataset2024.03 | 0.681 | 0.696 | |
| GRepQOU (Opinion-Unaware)=false, Train Dataset=PIPAL, Evaluation Setting=Cross-dataset2024.03 | 0.679 | 0.682 | |
| BRISQUETraining Dataset=KADID-10k, Evaluation=Zero-shot2026.04 | 0.665 | 0.681 | |
| CL-MISupervision type=Label-free2026.01 | 0.664 | 0.653 | |
| InternVL3.5-8BModel Category=Open-Source, Domain=Cross domain testing2025.12 | 0.663 | 0.66 | |
| CONTRIQUESupervision type=Label-free2026.01 | 0.651 | 0.637 |