Research map / Image generation / Text-to-image diffusion
PromptIQA: Boosting the Performance and Generalization for No-Reference Image Quality Assessment via Prompts
Zewen Chen, Haina Qin, Juan Wang, Chunfeng Yuan, Bing Li, Weiming Hu, Leon Wang
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 |
|---|---|---|---|
| Remove the image-score prompt components (w/o prompt) (ablation) | PromptIQA · Generalization TID2013 FSIM(MOS) | PLCC: 0.7851 -> 0.8802 | Improved |
| Add image-score prompt (ISPP) to mixed training (ablation) | PromptIQA · Generalization Kadid-10k LPIPS(DMOS) | PLCC: -0.7397 -> 0.7615 | Improved |
| Add random scaling and random flipping augmentations (ablation) | PromptIQA · Generalization Kadid-10k LPIPS(DMOS) | PLCC: 0.4960 -> 0.7615 | Improved |
| Add random scaling and flipping augmentations (ablation) | PromptIQA · Mixed test set TID2013 | PLCC: 0.9028 -> 0.9223 | Improved |
| Add random scaling and flipping augmentations (ablation) | PromptIQA · Mixed test set UWIQA | PLCC: 0.8791 -> 0.8839 | No clear effect |
| Train on the mixed 12 datasets instead of a single dataset (ablation) | PromptIQA · Generalization TID2013 LPIPS(DMOS) | PLCC: 0.6511 -> 0.8064 | Improved |
| Use interval sampling instead of random sampling for ISPs (ablation) | PromptIQA · Kadid-10k | SROCC: 0.9179 -> 0.9277 | Improved |
| Use random ISP selection on UWIQA (ablation) | PromptIQA · UWIQA | SROCC: 0.8766 -> 0.8773 | No clear effect |
| Replace images or scores in the prompt with random numbers (ablation) | PromptIQA · TID2013 | SROCC: 0.9030 -> 0.1287 | Hurt |
| Compared with the task-specific MoNet, use mixed-training PromptIQA on synthetic-distortion datasets | PromptIQA · SDN-IQA average | SROCC: 0.9263 -> 0.9231 | Hurt |
| Use 10 prompt images with no fine-tuning instead of fine-tuning baselines on the same 10 ISPs | PromptIQA vs MANIQA and MoNet · TID2013 with FSIM(MOS) labels | PLCC: 0.8422 -> 0.8802 | Improved |
| Increase number of ISPs in the prompt (3 to 10) (ablation) | PromptIQA · LIVEC, BID, SPAQ | performance | Improved |
Closest papers by meaning
- IQA-Spider: Unifying Multi-Granularity Image Quality Assessment with Reasoning, Grounding and Referring (ICML 2026 · similarity 0.85)
- Revisiting MLLM Based Image Quality Assessment: Errors and Remedy (AAAI 2026 · similarity 0.85)
- Causal-IQA: Towards the Generalization of Image Quality Assessment Based on Causal Inference (ICML 2024 · similarity 0.85)
- Adaptive Image Quality Assessment via Teaching Large Multimodal Model to Compare (NeurIPS 2024 · similarity 0.85)
- QMamba: On First Exploration of Vision Mamba for Image Quality Assessment (ICML 2025 · similarity 0.85)
- Grounding-IQA: Grounding Multimodal Language Model for Image Quality Assessment (ICLR 2026 · similarity 0.84)
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.
