Research map / Detection and segmentation / Open-set and few-shot detection
The Fishyscapes Benchmark: Measuring Blind Spots in Semantic Segmentation
Hermann Blum, Paul-Edouard Sarlin, Juan I. Nieto, Roland Siegwart, César Cadena
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 input noise preprocessing (ablation) | Learned Embedding Density · FS Static test | AP: 52.5 -> 55.4 | Improved |
| increase kNN neighbors from 1 to 20 (ablation) | kNN Embedding density · FS Lost & Found validation | AP: 42.3 -> 52.2 | Improved |
| increase kNN neighbors for class-based density from 5 to 100 (ablation) | kNN relative class density · FS Static validation | AP: 5.4 -> 9.9 | Improved |
| aggregate layers with minimum NLL instead of a single layer (ablation) | Learned Embedding Density · FS Static | AP: 40.9 -> 62.1 | Improved |
| aggregate layers with logistic regression instead of single layer (ablation) | Learned Embedding Density · FS Static | AP: 40.9 -> 57.2 | Improved |
| use minimum NLL instead of logistic regression aggregation (ablation) | Learned Embedding Density · FS Lost & Found | FPR95: 24.4 -> 47.2 | Hurt |
| use softmax entropy instead of max-probability (ablation) | Softmax baseline · FS Static | AP: 12.9 -> 15.4 | Improved |
| train with OoD data (void classifier) instead of no OoD training | DeepLabv3+ softmax baseline · FS Static | AP: 12.9 -> 45.0 | Improved |
| train with special OoD losses (Outlier Head) versus unmodified model | DeepLabv3+ · Cityscapes | mIoU: 80.3 -> 77.3 | Hurt |
| use Bayesian predictive entropy instead of mutual information | Bayesian DeepLab · FS Misclassification | AP: 70.4 -> 73.8 | Improved |
| use softmax max-probability for misclassification instead of anomaly detection | Softmax · FS Lost & Found vs FS Misclassification | AP: 1.8 -> 67.4 | Improved |
| add blended in-distribution objects to FS Web (Jan 20) | embedding-based methods · FS Web | AP | No clear effect |
Closest papers by meaning
- A Dataset for Semantic Segmentation in the Presence of Unknowns (CVPR 2025 · similarity 0.88)
- Self-Supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes (AAAI 2024 · similarity 0.88)
- Spotting the Unexpected (STU): A 3D LiDAR Dataset for Anomaly Segmentation in Autonomous Driving (CVPR 2025 · similarity 0.86)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.85)
- Predictive Uncertainty Quantification for Bird's Eye View Segmentation: A Benchmark and Novel Loss Function (ICLR 2025 · similarity 0.85)
- ValUES: A Framework for Systematic Validation of Uncertainty Estimation in Semantic Segmentation (ICLR 2024 · 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.
