No-Reference Image Quality Assessment on KonIQ-10k
0.967SROCCLGDM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LGDM2025.09 | 0.967 | 0.972 | — | |
| B_GatedActivation=Gated2025.09 | 0.953 | 0.962 | — | |
| B_Gated (Ours)Activation=Gated, Type=Non-diffusion2025.09 | 0.953 | 0.962 | — | |
| DP-IQA2025.09 | 0.942 | 0.951 | — | |
| DeQA-ScoreModel Category=VLM-based2025.09 | 0.941 | 0.953 | — | |
| Q-AlignModel Category=VLM-based2025.09 | 0.94 | 0.941 | — | |
| B_SigActivation=Sigmoid2025.09 | 0.938 | 0.947 | — | |
| B_Sig (Ours)Activation=Sigmoid, Type=Non-diffusion2025.09 | 0.938 | 0.947 | — | |
| QCNType=Non-diffusion2025.09 | 0.934 | 0.945 | — | |
| BActivation=Baseline2025.09 | 0.932 | 0.943 | — | |
| LoDAType=Non-diffusion2025.09 | 0.932 | 0.944 | — | |
| B (Baseline)Type=Non-diffusion2025.09 | 0.932 | 0.943 | — | |
| MUSIQModel Category=Discriminative Deep-Learning-based2025.09 | 0.929 | 0.924 | — | |
| TOPIQBackbone=ResNet502023.08 | 0.928 | 0.941 | — | |
| QPT-ResNet50backbone=ResNet502023.03 | 0.9271 | 0.9413 | — | |
| Ours-SBackbone=Swin-Tiny, Joint training (FR & NR tasks)=false, Architecture Type=Transformer-based2023.10 | 0.926 | 0.938 | — | |
| MAMIQA2026.02 | 0.926 | 0.937 | — | |
| MUSIQtraining=KonIQ-10k (Type B)2022.07 | 0.924 | 0.937 | — | |
| MUSIQ2023.08 | 0.924 | 0.937 | — | |
| KonCept512training=KonIQ-10k (Type B)2022.07 | 0.921 | 0.937 | — | |
| KonCept5122023.08 | 0.921 | 0.937 | — | |
| Ours-RBackbone=ResNet50, Joint training (FR & NR tasks)=false2023.10 | 0.917 | 0.925 | — | |
| MUSIQ2023.03 | 0.916 | 0.928 | — | |
| MUSIQType=Non-diffusion2025.09 | 0.916 | 0.928 | — | |
| MUSIQ2026.06 | 0.916 | 0.928 | — | |
| TRES2023.03 | 0.915 | 0.928 | — | |
| TRESArchitecture Type=CNN/Hand-crafted2023.10 | 0.915 | 0.928 | — | |
| TreS2026.02 | 0.915 | 0.918 | — | |
| TReSType=Non-diffusion2025.09 | 0.915 | 0.928 | — | |
| TReS2026.06 | 0.915 | 0.928 | — | |
| TRIQA2026.06 | 0.915 | 0.926 | — | |
| SLIDE-IQA2026.06 | 0.915 | 0.93 | — | |
| RE-IQAType=Non-diffusion2025.09 | 0.914 | 0.923 | — | |
| ReIQA2026.06 | 0.914 | 0.923 | — | |
| ResNet50backbone=ResNet502023.03 | 0.9121 | 0.927 | — | |
| MS-SCANet2026.02 | 0.909 | 0.903 | — | |
| Q-InsightModel Category=VLM-based2025.09 | 0.909 | 0.928 | — | |
| HyperIQA2023.03 | 0.9075 | 0.9205 | — | |
| HyperIQAArchitecture Type=CNN/Hand-crafted2023.10 | 0.906 | 0.917 | — | |
| HyperIQAType=Non-diffusion2025.09 | 0.906 | 0.917 | — | |
| HyperIQAModel Category=Discriminative Deep-Learning-based2025.09 | 0.906 | 0.917 | — | |
| HyperIQA2026.06 | 0.906 | 0.917 | — | |
| HyperIQAtraining=KonIQ-10k (Type B)2022.07 | 0.904 | 0.915 | — | |
| CONRTIQUE2023.03 | 0.896 | 0.901 | — | |
| UNIQUE2023.03 | 0.896 | 0.901 | — | |
| ReIQA-CBackbone=MoCov22026.06 | 0.896 | 0.912 | — | |
| CLIP-IQA+training=KonIQ-10k (Type B)2022.07 | 0.895 | 0.909 | — | |
| CLIP-IQA+Model Category=Discriminative Deep-Learning-based2025.09 | 0.895 | 0.909 | — | |
| VCRNettraining=KonIQ-10k (Type B)2022.07 | 0.894 | 0.909 | — | |
| CONTRIQUE2026.06 | 0.894 | 0.906 | — | |
| TIQAArchitecture Type=Transformer-based2023.10 | 0.892 | 0.903 | — | |
| TRIQ2026.02 | 0.892 | 0.903 | — | |
| TIQAType=Non-diffusion2025.09 | 0.892 | 0.903 | — | |
| SFA2023.03 | 0.8882 | 0.8966 | — | |
| MetaIQAArchitecture Type=CNN/Hand-crafted2023.10 | 0.887 | 0.856 | — | |
| MetaIQAType=Non-diffusion2025.09 | 0.887 | 0.856 | — | |
| SLIDE-IQA-CBackbone=DINOv32026.06 | 0.887 | 0.909 | — | |
| PQR2026.06 | 0.88 | 0.884 | — | |
| DB-CNN2023.03 | 0.878 | 0.8867 | — | |
| DBCNNArchitecture Type=CNN/Hand-crafted2023.10 | 0.875 | 0.884 | — | |
| DBCNN2026.02 | 0.875 | 0.884 | — | |
| DBCNNType=Non-diffusion2025.09 | 0.875 | 0.884 | — | |
| DBCNNModel Category=Discriminative Deep-Learning-based2025.09 | 0.875 | 0.884 | — | |
| DB-CNN2026.06 | 0.875 | 0.884 | — | |
| P2P-BMArchitecture Type=CNN/Hand-crafted2023.10 | 0.872 | 0.885 | — | |
| HyperIQA2026.02 | 0.872 | 0.885 | — | |
| P2P-BMType=Non-diffusion2025.09 | 0.872 | 0.885 | — | |
| PaQ-2-PiQ2023.03 | 0.87 | 0.88 | — | |
| PaQ-2-PiQ2026.06 | 0.87 | 0.88 | — | |
| ReIQA-Q2026.06 | 0.861 | 0.885 | — | |
| NIMAModel Category=Discriminative Deep-Learning-based2025.09 | 0.859 | 0.896 | — | |
| SLIDE-IQA-Q2026.06 | 0.854 | 0.869 | — | |
| VisualQuality-R1Model Category=VLM-based2025.09 | 0.852 | 0.882 | — | |
| Re-IQAModel Category=Discriminative Deep-Learning-based2025.09 | 0.838 | 0.87 | — | |
| MANIQAModel Category=Discriminative Deep-Learning-based2025.09 | 0.834 | 0.849 | — | |
| WaDIQaMArchitecture Type=CNN/Hand-crafted2023.10 | 0.804 | 0.807 | — | |
| WaDIQaMType=Non-diffusion2025.09 | 0.804 | 0.807 | — | |
| EvoQualityModel Category=VLM-based, Evolution Round=Round 12025.09 | 0.794 | 0.84 | — | |
| EvoQualityModel Category=VLM-based, Evolution Round=Round 22025.09 | 0.791 | 0.835 | — | |
| CORNIA2026.06 | 0.78 | 0.795 | — | |
| MM-IQA2026.04 | 0.769 | 0.785 | — | |
| TOPIQTrain on=SPAQ2023.08 | 0.7632 | 0.8112 | — | |
| TOPIQTrain on=FLIVE2023.08 | 0.7622 | 0.8008 | — | |
| Qwen2.5-VL-7BModel Category=VLM-based2025.09 | 0.761 | 0.703 | — | |
| CONTRIQUEModel Category=Discriminative Deep-Learning-based2025.09 | 0.754 | 0.798 | — | |
| NBIQA2026.04 | 0.749 | 0.771 | — | |
| ENIQA2026.04 | 0.745 | 0.761 | — | |
| SFD-IQA2025.03 | 0.744 | 0.7574 | 0.5443 | |
| MDFS2025.03 | 0.7333 | 0.7123 | 0.5344 | |
| WaDIQaM-NR2023.03 | 0.7294 | 0.7538 | — | |
| CurveletQA2026.04 | 0.718 | 0.73 | — | |
| BRISQUE2023.03 | 0.715 | 0.7016 | — | |
| MUSIQTrain on=FLIVE2023.08 | 0.7084 | 0.7655 | — | |
| TRESTrain on=FLIVE2023.08 | 0.7068 | 0.7507 | — | |
| BRISQUEtraining=None (Type A)2022.07 | 0.705 | 0.707 | — | |
| BRISQUE2023.08 | 0.705 | 0.707 | — | |
| GWH-GLBP2026.04 | 0.698 | 0.723 | — | |
| GM-LOG-BIQA2026.04 | 0.696 | 0.705 | — | |
| CLIP-IQAtraining=None (Type A)2022.07 | 0.695 | 0.727 | — | |
| DIIVINE2026.04 | 0.693 | 0.709 | — |