Research map / Image generation / Autoregressive and latent generation
From Global to Granular: Revealing IQA Model Performance via Correlation Surface
Baoliang Chen, Danni Huang, Hanwei Zhu, Lingyu Zhu, Wei Zhou, Shiqi Wang, Yuming Fang, Weisi Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add kernel-smoothed density estimation to the Distribution Regulator (ablation) | GMC · PIPAL (50 extreme sampling trials) | average std of GMC_g: 0.1664 -> 0.1651 | Improved |
| add kernel-smoothed density estimation to the Distribution Regulator (ablation) | GMC · SPAQ (50 extreme sampling trials) | average std of GMC_g: 0.1778 -> 0.1737 | Improved |
| add kernel-smoothed density estimation to the Distribution Regulator (ablation) | GMC · AIGCIQA2023 (50 extreme sampling trials) | average std of GMC_g: 0.1783 -> 0.1626 | Improved |
| use Latin Hypercube Sampling instead of random sampling for surface points (ablation) | GMC_g · PIPAL/SPAQ sampling size study | stability of GMC_g | Improved |
| select the best model with GMC_s instead of SRCC for high-quality retrieval | IQA model selection · SPAQ high-quality set | mean MOS of top-3: 76.37 -> 86.17 | Improved |
| select the best model with GMC_s instead of SRCC for high-quality retrieval | IQA model selection · KADID-10k high-quality set | mean MOS of top-3: 4.68 -> 4.75 | Improved |
| select the best model with GMC_s instead of SRCC for AIGC quality retrieval | IQA model selection · AGIQA-3K quality set | mean MOS of top-3: 3.48 -> 4.14 | Improved |
| integrate LPIPS with MS-SSIM (GMC-suggested) rather than DISTS | IQA model integration · PIPAL | SRCC: 0.6019 -> 0.6202 | Improved |
| use GMC_g instead of SRCC under shifted MOS distributions | evaluation metric · PIPAL and SPAQ nine sampled subsets | variability across subsets | Improved |
| optimize MS-SSIM instead of LPIPS or DISTS under adversarial constraint | fine-grained image optimization · PIPAL | visual quality of optimized images | Improved |
Closest papers by meaning
- Toward Generalized Image Quality Assessment: Relaxing the Perfect Reference Quality Assumption (CVPR 2025 · similarity 0.89)
- Adaptive Image Quality Assessment via Teaching Large Multimodal Model to Compare (NeurIPS 2024 · similarity 0.89)
- IQA-Spider: Unifying Multi-Granularity Image Quality Assessment with Reasoning, Grounding and Referring (ICML 2026 · similarity 0.88)
- Causal-IQA: Towards the Generalization of Image Quality Assessment Based on Causal Inference (ICML 2024 · similarity 0.88)
- Revisiting MLLM Based Image Quality Assessment: Errors and Remedy (AAAI 2026 · similarity 0.87)
- Image Quality Assessment: From Human to Machine Preference (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
