Research map / Probabilistic and scientific modeling / Differential equations and PDEs
PIE: Physics-Inspired Low-Light Enhancement
Dong Liang, Zhengyan Xu, Ling Li, Mingqiang Wei, Songcan Chen
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 contrastive loss L_c (ablation) | PIE · Average of five test sets | NIQE: 4.38 -> 3.38 | Improved |
| add regional brightness consistency loss L_rc (ablation) | PIE · Average of five test sets | NIQE: 3.45 -> 3.38 | Improved |
| add feature retention loss L_fr (ablation) | PIE · Average of five test sets | NIQE: 3.53 -> 3.38 | Improved |
| add negative samples (ablation) | PIE · Average of five test sets | UNIQUE: 0.78 -> 0.95 | Improved |
| add overexposed negatives (ablation) | PIE · Average of five test sets | UNIQUE: 0.74 -> 0.95 | Improved |
| add underexposed negatives (ablation) | PIE · Average of five test sets | NIQE: 4.38 -> 3.38 | Improved |
| use mixed Bag of Curves instead of gamma curves only (ablation) | PIE · Average of five test sets | NIQE: 3.39 -> 3.38 | No clear effect |
| use mixed Bag of Curves instead of logarithmic curves only (ablation) | PIE · Average of five test sets | NIQE: 3.47 -> 3.38 | Improved |
| use InfoNCE instead of triplet loss for contrastive learning (ablation) | PIE · SICE Part2 | PSNR: 18.44 -> 19.79 | Improved |
| use InfoNCE instead of N-pair loss (ablation) | PIE · SICE Part2 | PSNR: 15.08 -> 19.79 | Improved |
| use 5 negatives per positive instead of 1 (ablation) | PIE · Average of five test sets | NIQE: 3.38 -> 3.37; time 73.6 -> 131.3 min | No clear effect |
| use 5 positives per negative instead of 1 (ablation) | PIE · Average of five test sets | NIQE: 3.38 -> 3.46 | Hurt |
Closest papers by meaning
- Semantically Contrastive Learning for Low-Light Image Enhancement (AAAI 2022 · similarity 0.89)
- Zero-Reference Low-Light Enhancement via Physical Quadruple Priors (CVPR 2024 · similarity 0.86)
- Luminance-Aware Statistical Quantization: Unsupervised Hierarchical Learning for Illumination Enhancement (NeurIPS 2025 · similarity 0.86)
- Zero-Shot Low-Light Image Enhancement via Latent Diffusion Models (AAAI 2025 · similarity 0.86)
- Improving Visual and Downstream Performance of Low-Light Enhancer with Vision Foundation Models Collaboration (CVPR 2025 · similarity 0.85)
- FRBNet: Revisiting Low-Light Vision through Frequency-Domain Radial Basis Network (NeurIPS 2025 · similarity 0.85)
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.
