Research map / Image generation / Autoregressive and latent generation
An Unsupervised Information-Theoretic Perceptual Quality Metric
Sangnie Bhardwaj, Ian Fischer, Johannes Ballé, Troy Chinen
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 |
|---|---|---|---|
| use PIM instead of LPIPS Alex-lin | perceptual quality metric evaluation · BAPPS-JND All | mAP (%): 61.50 -> 64.40 | Improved |
| use PIM instead of LPIPS Alex-lin | compression method ranking prediction · CLIC 2020 | Spearman's ρ: --0.847 -> --0.864 | Improved |
| replace IXYZ with InfoNCE (ablation) | PIM · BAPPS-2AFC | human agreement (%) | Hurt |
| replace IXYZ with InfoNCE (ablation) | PIM · BAPPS-JND | mAP (%) | Hurt |
| replace IXYZ with single variable InfoNCE (ablation) | PIM · BAPPS-2AFC | human agreement (%): 69.11% -> 65.15% | Hurt |
| replace IXYZ with single variable InfoNCE (ablation) | PIM · BAPPS-JND | mAP (%): 64.40% -> 44.79% | Hurt |
| collapse the marginal encoders to one mixture component (ablation) | PIM · BAPPS-2AFC All | human agreement (%): 69.11 -> 68.82 | Hurt |
| collapse the marginal encoders to one mixture component (ablation) | PIM · BAPPS-JND All | mAP (%): 64.40 -> 64.46 | No clear effect |
| remove the frontend multi-scale decomposition (ablation) | PIM · BAPPS | human-rating prediction performance | Hurt |
| replace the steerable pyramid with a Laplacian pyramid or shallower frontend CNN (ablation) | PIM · BAPPS | human-rating prediction performance | Hurt |
| shift reference images by a few pixels | PIM · BAPPS-JND, 2-pixel shift | score change: --0.97 | Hurt |
Closest papers by meaning
- The Unreasonable Effectiveness of Linear Prediction as a Perceptual Metric (ICLR 2024 · similarity 0.88)
- Multidimensional Observer Model and Perceptual Dimensions of Human Image Quality Assessment (NeurIPS 2026 · similarity 0.88)
- Augmenting Perceptual Super-Resolution via Image Quality Predictors (CVPR 2025 · similarity 0.86)
- PerceptionRubrics: Calibrating Multimodal Evaluation to Human Perception (ICML 2026 · similarity 0.86)
- FoMo: Forking Moment in Generative Trajectory as a Perceptual Distance (NeurIPS 2026 · similarity 0.86)
- Evaluation of Visual Processing Should Be Human-Centered, Not Metric-Centered (NeurIPS 2026 · 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.
