Research map / Image generation / Autoregressive and latent generation
Detecting Road Obstacles by Erasing Them
Krzysztof Lis, Sina Honari, Pascal Fua, Mathieu Salzmann
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 inpainting of road patches (vs none) (ablation) | Ours · RoadObstacles21 (validation) | instance F1: 54.7 -> 60.0 | Improved |
| add inpainting (vs none) (ablation) | Ours · RoadObstacles21 (validation) | pixel AP: 89.9 -> 91.0 | Improved |
| add discrepancy network (vs L1 distance) (ablation) | Ours · Lost and Found (test no known) | pixel AP: 19.6 -> 82.3 | Improved |
| add inpainting instead of resynthesis (ablation) | Ours · RoadObstacles21 (validation) | instance F1: 45.2 -> 60.0 | Improved |
| use inpainting instead of Gaussian blur (ablation) | Ours · RoadObstacles21 (validation) | instance F1: 55.3 -> 60.0 | Improved |
| add road segmentation plus detector instead of segmentation alone (ablation) | Ours · RoadObstacles21 (validation) | instance F1: 26.1 -> 60.0 | Improved |
| add input Gaussian blur (ablation) | Ours · RoadObstacles21 (validation) | pixel AP: 87.7 -> 91.0 | Improved |
| add input Gaussian blur (ablation) | Ours · Lost and Found (test no known) | instance F1: 56.2 -> 54.0 | Hurt |
| use ground-truth road area instead of segmented road area | Ours · Lost and Found (test no known) | FPR95: 68.5 -> 2.3 | Improved |
| use ground-truth road area instead of segmented road area | Ours · RoadObstacles21 (test) | pixel AP: 67.5 -> 67.8 | No clear effect |
Closest papers by meaning
- Beyond Pixel Uncertainty: Bounding the OoD Objects in Road Scenes (ICCV 2025 · similarity 0.77)
- Road Anomaly Detection by Partial Image Reconstruction With Segmentation Coupling (ICCV 2021 · similarity 0.76)
- Instruct2See: Learning to Remove Any Obstructions Across Distributions (ICML 2025 · similarity 0.76)
- 3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline (ICLR 2025 · similarity 0.75)
- Beware of Road Markings: A New Adversarial Patch Attack to Monocular Depth Estimation (NeurIPS 2024 · similarity 0.75)
- GeoRemover: Removing Objects and Their Causal Visual Artifacts (NeurIPS 2025 · similarity 0.74)
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.
