Research map / Detection and segmentation / Open-set and few-shot detection
Physics-Based Rendering for Improving Robustness to Rain
Shirsendu Sukanta Halder, Jean-Francois Lalonde, Raoul de Charette
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 |
|---|---|---|---|
| finetune Faster R-CNN with curriculum on rendered rain (25 to 100 mm/hr) | Faster R-CNN · Augmented Kitti, 100mm/hr | mAP: 60.9% -> 64.5% | Improved |
| finetune Faster R-CNN with curriculum on rendered rain | Faster R-CNN · Augmented Kitti, 200mm/hr (unseen) | mAP: 55.4% -> 62.4% | Improved |
| finetune PSPNet with curriculum on rendered rain | PSPNet · Augmented Cityscapes, 100mm/hr | AP: 52.0% -> 54.0% | Improved |
| finetune Faster R-CNN on rendered rain, evaluate on real rain | Faster R-CNN · nuScenes rainy | mAP: +14.9% | Improved |
| finetune PSPNet on rendered rain, evaluate on real rain | PSPNet · nuScenes rainy | AP: +36.6% | Improved |
| finetune on rendered rain, evaluate in clear weather | Faster R-CNN and PSPNet · nuScenes clear | mAP / AP | No clear effect |
| train with rain from scratch instead of refining a clear-weather model (ablation) | Faster R-CNN and PSPNet · synthetic rain | convergence | Hurt |
| increase rain to 200mm/hr relative to clear weather | six KITTI object detectors · Augmented Kitti | mAP relative drop: at least 12% worse, 25-30% for R-FCN, SSD, MX-RCNN | Hurt |
| strong rain (200mm/hr) vs clear | six Cityscapes segmentation models · Augmented Cityscapes | AP relative drop: almost 60% under strong rain | Hurt |
| strong rain (200mm/hr) vs clear | ESPNet · Augmented Cityscapes | AP relative drop: 85% | Hurt |
Closest papers by meaning
- Sunshine to Rainstorm: Cross-Weather Knowledge Distillation for Robust 3D Object Detection (AAAI 2024 · similarity 0.88)
- RainyGS: Efficient Rain Synthesis with Physically-Based Gaussian Splatting (CVPR 2025 · similarity 0.88)
- Semantic Segmentation on Raindrop Degraded Images Using Two-Stage Dual Teacher-Student Learning (AAAI 2025 · similarity 0.87)
- Towards Universal Rainy Image Restoration: Benchmark and Baseline (AAAI 2025 · similarity 0.87)
- Harnessing Joint Rain-/Detail-aware Representations to Eliminate Intricate Rains (ICLR 2024 · similarity 0.87)
- ERF: A Benchmark Dataset for Robust Semantic Segmentation Under Extreme Rainfall Conditions (AAAI 2025 · similarity 0.87)
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.
