Research map / Detection and segmentation / Open-set and few-shot detection
Random Walk on Pixel Manifolds for Anomaly Segmentation of Complex Driving Scenes
Zelong Zeng, Kaname Tomite
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 RWPM random-walk embedding refinement at inference | PEBAL · Fishyscapes Lost&Found | AP: 58.81 -> 66.85 | Improved |
| add RWPM | Balanced Energy · Road Anomaly | FPR95: 41.46 -> 32.10 | Improved |
| add RWPM | Mask2Anomaly · Fishyscapes Lost&Found | AP: 65.27 -> 65.38 | No clear effect |
| add RWPM | RbA · SMIYC Anomaly Track component-level | mean F1: 46.80 -> 58.44 | Improved |
| partition embedding map into n x n sub-maps instead of full map (ablation) | RWPM · Road Anomaly | FPS: 0.32 -> 2.04 FPS, AP 87.90 -> 87.34 | Improved |
| add sub-map calibration when n>2 (ablation) | RWPM · Road Anomaly | FPR95: 6.31 -> 5.32 (n=4) | Improved |
| add calibration at n=2 (ablation) | RWPM · Road Anomaly | AP: 87.34 -> 87.18 | No clear effect |
| use limited iterations instead of closed form (T=20 vs infinity) (ablation) | RWPM · Road Anomaly | AP: 87.51 -> 87.34, FPS 0.29 -> 2.04 | No clear effect |
| increase iteration number T (ablation) | RWPM · Road Anomaly | AP: 86.89 (T=5) -> 87.54 (T=100), FPS 3.70 -> 0.50 | Mixed |
| reduce embeddings by resizing or superpixels instead of partitioning (ablation) | RWPM · Road Anomaly | AP: memory 23.82GiB (resize 0.9) vs 7.24GiB | Hurt |
| add RWPM to RbA for in-distribution segmentation | RbA · Cityscapes val | mIoU: 81.93 -> 82.16 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Mask2Anomaly — Mask2Anomaly: Mask Transformer for Universal Open-Set Segmentation (TPAMI 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Spotting the Unexpected (STU): A 3D LiDAR Dataset for Anomaly Segmentation in Autonomous Driving (CVPR 2025 · similarity 0.86)
- Self-Supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes (AAAI 2024 · similarity 0.84)
- A Dataset for Semantic Segmentation in the Presence of Unknowns (CVPR 2025 · similarity 0.83)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.83)
- Manifold Induced Biases for Zero-shot and Few-shot Detection of Generated Images (ICLR 2025 · similarity 0.83)
- Odd-One-Out: Anomaly Detection by Comparing with Neighbors (CVPR 2025 · similarity 0.82)
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.
