No-reference Image Quality Assessment on LIVEW
91.4PLCCQ-DeepSight
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Q-DeepSightCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 91.4 | 89.1 | |
| Q-DeepSight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 91.4 | 89.3 | |
| Q-Insight†Category=MLLM (w/ reasoning), Training Dataset=KonIQ + DQ-7K2026.04 | 89.3 | 86.5 | |
| DeQA-ScoreCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 89.2 | 87.9 | |
| DeQA-ScoreModel Type=Non-reasoning2026.03 | 89.1 | 87.1 | |
| DeQA-ScoreModel Category=VLM-based, Zero-shot evaluation=true2025.09 | 89.1 | 87.1 | |
| ME-IQABase VLM=EvoQuality2026.03 | 88.7 | 86 | |
| Zoom-IQACategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 88.7 | 87 | |
| ME-IQABase VLM=Q-Insight2026.03 | 88.1 | 84.2 | |
| ME-IQABase VLM=Doubao-Seed-1.6-vision2026.03 | 87.2 | 84.6 | |
| Q-InsightCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 87 | 83.9 | |
| Q-InsightModel Category=VLM-based, Zero-shot evaluation=true2025.09 | 87 | 83.9 | |
| ME-IQABase VLM=VisualQuality-R12026.03 | 86.8 | 85.3 | |
| BaselineBase VLM=EvoQuality2026.03 | 86 | 82.9 | |
| ME-IQABase VLM=GPT-52026.03 | 86 | 83.2 | |
| BaselineBase VLM=Q-Insight2026.03 | 85.9 | 81.7 | |
| BaselineBase VLM=Doubao-Seed-1.6-vision2026.03 | 85.8 | 83.7 | |
| Compare2ScoreModel Type=Non-reasoning2026.03 | 85.7 | 82.1 | |
| VisualQuality-R1Category=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 85.6 | 82.7 | |
| VisualQuality-R1Model Category=VLM-based, Zero-shot evaluation=true2025.09 | 85.6 | 82.7 | |
| Q-AlignModel Type=Non-reasoning2026.03 | 85.3 | 86 | |
| Q-AlignCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 85.3 | 86 | |
| Q-AlignModel Category=VLM-based, Zero-shot evaluation=true2025.09 | 85.3 | 86 | |
| BaselineBase VLM=GPT-52026.03 | 85.1 | 82.3 | |
| MANIQAModel Type=Non-reasoning2026.03 | 84.9 | 83.2 | |
| ManIQACategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 84.9 | 83.2 | |
| MANIQAModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 84.9 | 83.2 | |
| BaselineBase VLM=VisualQuality-R12026.03 | 84.7 | 83.1 | |
| EvoQualityModel Category=VLM-based, Zero-shot evaluation=true, Evolution Round=Round 12025.09 | 84.7 | 81.4 | |
| EvoQualityModel Category=VLM-based, Zero-shot evaluation=true, Evolution Round=Round 22025.09 | 84.7 | 81.3 | |
| CLIP-IQA+Model Type=Non-reasoning2026.03 | 83.2 | 80.5 | |
| CLIP-IQA+Category=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 83.2 | 80.5 | |
| CLIP-IQA+Model Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 83.2 | 80.5 | |
| NIMAModel Type=Non-reasoning2026.03 | 81.4 | 77.1 | |
| NIMAModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 81.4 | 77.1 | |
| MUSIQModel Type=Non-reasoning2026.03 | 78.9 | 83 | |
| MUSIQCategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 78.9 | 83 | |
| MUSIQModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 78.9 | 83 | |
| C2ScoreCategory=MLLM (w/o reasoning), Training Dataset=KonIQ2026.04 | 78.6 | 77.2 | |
| DBCNNCategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 77.3 | 75.5 | |
| DBCNNModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 77.3 | 75.5 | |
| HyperIQACategory=Non-MLLM, Training Dataset=KonIQ2026.04 | 77.2 | 74.9 | |
| HyperIQAModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 77.2 | 74.9 | |
| Qwen-SFTCategory=MLLM (w/ reasoning), Training Dataset=KonIQ2026.04 | 73.4 | 72.8 | |
| Qwen2.5-VL-7BModel Category=VLM-based, Zero-shot evaluation=true2025.09 | 71.8 | 70 | |
| CONTRIQUEModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 71.3 | 67.6 | |
| Re-IQAModel Category=Discriminative Deep-Learning-based, Zero-shot evaluation=true2025.09 | 67.1 | 62.4 | |
| NIQEModel Category=Handcrafted, Zero-shot evaluation=true2025.09 | 49.3 | 44.9 | |
| NIQEModel Type=Non-reasoning2026.03 | 44.9 | 44.9 | |
| BRISQUEModel Category=Handcrafted, Zero-shot evaluation=true2025.09 | 36.1 | 31.3 |