No-Reference Image Quality Assessment on KADID (test)
0.934SROCCCONTRIQUE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CONTRIQUEModel Type=Unsupervised pretraining and Linear Regression2021.10 | 0.934 | 0.937 | |
| HyperIQAModel Type=Supervised pretraining and supervised fine-tuning2021.10 | 0.852 | 0.845 | |
| DB-CNNModel Type=Supervised pretraining and supervised fine-tuning2021.10 | 0.851 | 0.856 | |
| OursCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.734 | 0.672 | |
| Q-Insight-ScoreModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.702 | 0.702 | |
| Q-Insight-ScoreCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.702 | 0.702 | |
| Resnet-50Model Type=Supervised pretraining and Linear Regression2021.10 | 0.701 | 0.677 | |
| Autoencoder-likeModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.7 | 0.649 | |
| Chain-of-ThoughtModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.699 | 0.629 | |
| Self-ConsistencyModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.692 | 0.632 | |
| DeQAModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.687 | 0.694 | |
| DeQACategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.687 | 0.694 | |
| Q-AlignModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.684 | 0.674 | |
| Q-AlignCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.684 | 0.674 | |
| OursCategory=Caption-Only Conditions, Condition=Caption-Only2025.12 | 0.676 | 0.662 | |
| Self-ConsistencyInput Condition=Text-Only2026.01 | 0.661 | 0.627 | |
| Self-ConsistencyInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.66 | 0.632 | |
| CLIP-IQA+Model category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.654 | 0.653 | |
| CLIP-IQA+Category=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.654 | 0.653 | |
| Q-Insight-ScoreCategory=Caption-Only Conditions, Condition=Caption-Only2025.12 | 0.65 | 0.645 | |
| Autoencoder-likeInput Condition=Text-Only2026.01 | 0.645 | 0.632 | |
| Autoencoder-likeInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.644 | 0.638 | |
| Q-Insight-ScoreInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.621 | 0.611 | |
| Chain-of-ThoughtInput Condition=Text-Only, Text Filtering=Score-related words removed2026.01 | 0.621 | 0.609 | |
| Q-Insight-ScoreInput Condition=Text-Only2026.01 | 0.62 | 0.604 | |
| Chain-of-ThoughtInput Condition=Text-Only2026.01 | 0.62 | 0.604 | |
| HOSAModel Type=Codebook-based Features2021.10 | 0.618 | 0.653 | |
| MUSIQModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.556 | 0.575 | |
| MUSIQCategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.556 | 0.575 | |
| NIMAModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.535 | 0.532 | |
| NIMACategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.535 | 0.532 | |
| BRISQUEModel Type=Traditional/Handcrafted Features2021.10 | 0.528 | 0.567 | |
| CORNIAModel Type=Codebook-based Features2021.10 | 0.516 | 0.558 | |
| DBCNNModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.484 | 0.497 | |
| DBCNNCategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.484 | 0.497 | |
| MANIQAModel category=Deep-Learning Models, Input Condition=Image-conditioned2026.01 | 0.465 | 0.499 | |
| MANIQACategory=Deep-Learning Models, Condition=Image-conditioned2025.12 | 0.465 | 0.499 | |
| C2ScoreModel category=SFT-based and RL-based MLLMs, Input Condition=Image-conditioned2026.01 | 0.453 | 0.5 | |
| C2ScoreCategory=SFT-based and RL-based MLLMs, Condition=Image-conditioned2025.12 | 0.453 | 0.5 | |
| NIQECategory=Hand-Crafted Models, Condition=Image-conditioned2025.12 | 0.405 | 0.468 | |
| NIQEModel Type=Traditional/Handcrafted Features2021.10 | 0.374 | 0.428 | |
| BRISQUECategory=Hand-Crafted Models, Condition=Image-conditioned2025.12 | 0.356 | 0.429 |