Research map / Detection and segmentation / CNN features
Standardized Max Logits: A Simple yet Effective Approach for Identifying Unexpected Road Obstacles in Urban-Scene Segmentation
Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, Jaegul Choo
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 |
|---|---|---|---|
| standardize max logits class-wise (SML) instead of raw max logit | Max Logit · Fishyscapes Lost & Found (validation) | AP: 18.77 -> 27.61 | Improved |
| standardize max logits class-wise (SML) instead of raw max logit | Max Logit · Fishyscapes Lost & Found (validation) | FPR95: 38.13 -> 15.46 | Improved |
| add iterative boundary suppression (ablation) | SML · Fishyscapes Lost & Found (validation) | AP: 27.61 -> 31.63 | Improved |
| add iterative boundary suppression (ablation) | SML · Fishyscapes Lost & Found (validation) | FPR95: 15.46 -> 14.58 | Improved |
| add dilated smoothing without boundary suppression (ablation) | SML · Fishyscapes Lost & Found (validation) | AP: 27.61 -> 36.00 | Improved |
| add dilated smoothing without boundary suppression (ablation) | SML · Fishyscapes Lost & Found (validation) | FPR95: 15.46 -> 15.65 | Hurt |
| add dilated smoothing after boundary suppression (ablation) | SML + B Supp. · Fishyscapes Lost & Found (validation) | AP: 31.63 -> 36.55 | Improved |
| apply boundary suppression and dilated smoothing (ablation) | MSP · Fishyscapes Lost & Found (validation) | AP: 1.08 | No clear effect |
| apply boundary suppression and dilated smoothing (ablation) | Max Logit · Fishyscapes Lost & Found (validation) | AP: -1.45 | Hurt |
| apply boundary suppression and dilated smoothing (ablation) | SML · Fishyscapes Lost & Found (validation) | AP: 8.95 | Improved |
| increase boundary suppression iterations from 1 to 4 (ablation) | SML + B Supp. + D. Smoothing · Fishyscapes Lost & Found (validation) | AP: 36.26 -> 36.55 | Improved |
| increase boundary suppression iterations from 4 to 5 (ablation) | SML + B Supp. + D. Smoothing · Fishyscapes Lost & Found (validation) | AP: 36.55 -> 36.44 | No clear effect |
Closest papers by meaning
- SegmentMeIfYouCan: A Benchmark for Anomaly Segmentation (NeurIPS 2021 · similarity 0.89)
- R$$^{2}$$S100K: Road-Region Segmentation Dataset for Semi-supervised Autonomous Driving in the Wild (IJCV 2024 · similarity 0.87)
- Full-Resolution Residual Networks for Semantic Segmentation in Street Scenes (CVPR 2017 · similarity 0.86)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.86)
- Self-Supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes (AAAI 2024 · similarity 0.85)
- Spotting the Unexpected (STU): A 3D LiDAR Dataset for Anomaly Segmentation in Autonomous Driving (CVPR 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.
