Image Quality Assessment on KonIQ
0.951SRCCQ-Hawkeye
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Q-HawkeyeTraining Dataset=KonIQ2026.01 | 0.951 | 0.959 | |
| Q-HawkeyeMethodology=MLLM-Based Methods2026.01 | 0.951 | 0.959 | |
| DeQA-ScoreTraining Dataset=KonIQ, SPAQ, KADID, PIPAL2026.01 | 0.946 | 0.958 | |
| DeQA-ScoreTraining Dataset=KonIQ, SPAQ, KADID2026.01 | 0.944 | 0.957 | |
| DeQA-ScoreArchitecture=VLM-based, Training=Co-trained on KONIQ, SPAQ, and KADID2025.09 | 0.944 | 0.957 | |
| DeQA-ScoreTraining Dataset=KonIQ, SPAQ2026.01 | 0.943 | 0.953 | |
| DP-IQAModel Type=teacher2024.05 | 0.942 | 0.951 | |
| Q-DeepSightCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.942 | 0.953 | |
| DeQA-ScoreVenue=CVPR’25, Methodology=MLLM-Based Methods2026.01 | 0.941 | 0.953 | |
| DeQACategory=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.941 | 0.953 | |
| DeQACategory=MLLM-based, Training Dataset=KonIQ2025.10 | 0.941 | 0.953 | |
| DeQA-ScoreCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 0.941 | 0.953 | |
| Q-AlignVenue=ICML’24, Methodology=MLLM-Based Methods2026.01 | 0.94 | 0.941 | |
| Q-AlignCategory=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.94 | 0.941 | |
| Q-AlignCategory=MLLM-based, Training Dataset=KonIQ2025.10 | 0.94 | 0.941 | |
| Q-Align2024.05 | 0.94 | 0.941 | |
| Q-AlignCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 0.94 | 0.941 | |
| Q-AlignArchitecture=VLM-based, Training=Co-trained on KONIQ, SPAQ, and KADID2025.09 | 0.938 | 0.945 | |
| Q-DeepSight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 0.936 | 0.949 | |
| LoDa2024.11 | 0.932 | 0.944 | |
| LoDa2024.05 | 0.932 | 0.944 | |
| SaTQA2024.05 | 0.93 | 0.941 | |
| MUSIQVenue=ICCV’21, Methodology=Non-MLLM Deep-Learning2026.01 | 0.929 | 0.924 | |
| MUSIQCategory=Non-VLM, Trained on KonIQ=true2026.01 | 0.929 | 0.924 | |
| MUSIQCategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.929 | 0.924 | |
| MUSIQCategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 0.929 | 0.924 | |
| LIQE2024.05 | 0.928 | 0.912 | |
| DP-IQAModel Type=student2024.05 | 0.926 | 0.944 | |
| CSFIQA2024.11 | 0.924 | 0.944 | |
| Zoom-IQACategory=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.922 | 0.938 | |
| QFM-IQM2024.11 | 0.922 | 0.936 | |
| RALICategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.922 | 0.939 | |
| Zoom-IQACategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.922 | 0.938 | |
| DEIQT2024.11 | 0.921 | 0.934 | |
| DEIQT2024.05 | 0.921 | 0.934 | |
| Q-insightTraining Dataset=KonIQ, KADIS2026.01 | 0.916 | 0.933 | |
| MUSIQ2024.11 | 0.916 | 0.928 | |
| CDINet2024.11 | 0.916 | 0.928 | |
| Q-InsightCategory=MLLM-based, Training Dataset=KonIQ2025.10 | 0.916 | 0.933 | |
| MUSIQ2024.05 | 0.916 | 0.928 | |
| Q-Insight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 0.916 | 0.933 | |
| TReS2024.11 | 0.915 | 0.928 | |
| TReS2024.05 | 0.915 | 0.928 | |
| Re-IQA2024.11 | 0.914 | 0.923 | |
| ReIQA2024.05 | 0.914 | 0.923 | |
| Compare2ScoreVenue=NIPS’24, Methodology=MLLM-Based Methods2026.01 | 0.91 | 0.923 | |
| C2ScoreCategory=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.91 | 0.923 | |
| C2ScoreCategory=MLLM-based, Training Dataset=KonIQ2025.10 | 0.91 | 0.923 | |
| C2ScoreCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 0.91 | 0.923 | |
| VisualQuality-R1Venue=NeurIPS’25, Methodology=MLLM-Based Methods2026.01 | 0.908 | 0.921 | |
| VisualQuality-R1Category=MLLM-based, Training Dataset=KonIQ2025.10 | 0.908 | 0.923 | |
| HyperIQAVenue=CVPR’20, Methodology=Non-MLLM Deep-Learning2026.01 | 0.906 | 0.917 | |
| HyperIQACategory=Non-VLM, Trained on KonIQ=true2026.01 | 0.906 | 0.917 | |
| HyperIQA2024.11 | 0.906 | 0.917 | |
| HyperIQA2024.05 | 0.906 | 0.917 | |
| HyperIQACategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 0.906 | 0.917 | |
| CLIP-IQA+Venue=AAAI’23, Methodology=Non-MLLM Deep-Learning2026.01 | 0.905 | 0.909 | |
| DACNN2024.11 | 0.901 | 0.912 | |
| VisualQuality-R1Architecture=VLM-based, Training=Co-trained on KONIQ, SPAQ, and KADID2025.09 | 0.899 | 0.915 | |
| VQ-R1+EvoQualityArchitecture=VLM-based, Training=Co-trained on KONIQ, SPAQ, and KADID2025.09 | 0.899 | 0.915 | |
| VisualQuality-R1Category=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.896 | 0.91 | |
| VisualQuality-R1Category=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.896 | 0.91 | |
| Q-insightVenue=NeurIPS’25, Methodology=MLLM-Based Methods2026.01 | 0.895 | 0.918 | |
| CLIP-IQA+Category=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.895 | 0.909 | |
| Q-InsightCategory=VLM (w/o & w/ reasoning), Trained on KonIQ=true2026.01 | 0.895 | 0.918 | |
| CLIP-IQA+2024.11 | 0.895 | 0.909 | |
| LIQECategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.895 | 0.901 | |
| CLIP-IQA2024.05 | 0.895 | 0.909 | |
| CLIP-IQA+Category=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 0.895 | 0.909 | |
| Q-InsightCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.895 | 0.918 | |
| TIQA2024.05 | 0.892 | 0.903 | |
| MetaIQA2024.11 | 0.887 | 0.856 | |
| MetaIQA2024.05 | 0.887 | 0.856 | |
| DBCNNVenue=ICSIPA’19, Methodology=Non-MLLM Deep-Learning2026.01 | 0.875 | 0.884 | |
| DBCNNCategory=Non-VLM, Trained on KonIQ=true2026.01 | 0.875 | 0.884 | |
| DBCNN2024.11 | 0.875 | 0.884 | |
| DBCNN2024.05 | 0.875 | 0.884 | |
| DBCNNCategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 0.875 | 0.884 | |
| VisualQuality-R1Training Dataset=KADID, SPAQ2026.01 | 0.874 | 0.89 | |
| P2P-BM2024.11 | 0.872 | 0.885 | |
| P2P-BM2024.05 | 0.872 | 0.885 | |
| Q-ProbeCategory=MLLMs-based2026.01 | 0.871 | 0.863 | |
| Qwen-SFTVenue=arXiv’25, Methodology=MLLM-Based Methods2026.01 | 0.866 | 0.889 | |
| Qwen-SFTCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 0.866 | 0.889 | |
| NIMAVenue=TIP’18, Methodology=Non-MLLM Deep-Learning2026.01 | 0.859 | 0.896 | |
| NIMACategory=Non-VLM, Trained on KonIQ=true2026.01 | 0.859 | 0.896 | |
| VisualQuality-R1Category=MLLMs-based2026.01 | 0.855 | 0.84 | |
| ManIQACategory=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.849 | 0.895 | |
| MG-IQAmulti-dataset training=true, training_data=KADID-10k + SPAQ2026.04 | 0.848 | 0.872 | |
| ManIQAVenue=CVPR’22, Methodology=Non-MLLM Deep-Learning2026.01 | 0.834 | 0.849 | |
| ManIQACategory=Non-VLM, Trained on KonIQ=true2026.01 | 0.834 | 0.849 | |
| Q-Aligninference time=0.1603s2025.12 | 0.834 | 0.7086 | |
| CLIP-IQA+Category=Non-MLLM Deep-learning, Training Dataset=KonIQ2025.10 | 0.834 | 0.909 | |
| ManIQACategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 0.834 | 0.849 | |
| VQ-R1multi-dataset training=true, training_data=KADID-10k + SPAQ2026.04 | 0.828 | 0.853 | |
| Tool-IQATraining dataset=KADID-10K, Method category=VLM-based2026.06 | 0.825 | 0.848 | |
| UnifiedReward-TCategory=MLLMs-based2026.01 | 0.82 | 0.804 | |
| Q-Insightmulti-dataset training=true, training_data=KADID-10k + SPAQ2026.04 | 0.811 | 0.836 | |
| Q-InsightCategory=MLLMs-based2026.01 | 0.806 | 0.829 | |
| QInsightTraining dataset=KADID-10K, Method category=VLM-based2026.06 | 0.806 | 0.829 |