Research map / Label-efficient and robust learning / Recognition and long-tail
WildDash - Creating Hazard-Aware Benchmarks
Oliver Zendel, Katrin Honauer, Markus Murschitz, Daniel Steininger, Gustavo Fernandez Dominguez
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 |
|---|---|---|---|
| train on Mapillary subset instead of Cityscapes (ablation) | DRN-d38 · WildDash val | mIoU: 16.5 -> 29.3 | Improved |
| train on all public Mapillary data instead of a similar-sized subset (ablation) | DRN-d38 · WildDash val | mIoU: 29.3 -> 30.7 | Improved |
| train on all public Mapillary data instead of a subset (ablation) | DRN-d38 · WildDash negative test cases | mIoU: 12.9 -> 27.4 | Improved |
| train on all public Mapillary data instead of a subset (ablation) | DRN-d38 · Cityscapes val | mIoU: 44.81 -> 46.34 | Improved |
| train on Mapillary subset instead of Cityscapes (ablation) | DRN-d38 · Cityscapes val | mIoU: 63.79 -> 44.81 | Hurt |
| increase over-exposure from none to high severity | mapillary+ · WildDash over-exposure subset | mIoU: 32.2 -> 17.0 | Hurt |
| increase under-exposure from none to high severity | mapillary+ · WildDash under-exposure subset | mIoU: 31.5 -> 20.8 | Hurt |
| increase hood visibility from none to high | mapillary+ · WildDash hood subset | mIoU: 32.9 -> 22.4 | Hurt |
| increase blur from none to high | mapillary+ · WildDash blur subset | mIoU: 29.0 -> 26.6 | Hurt |
| increase lens distortions from none to high | mapillary+ · WildDash distortions subset | mIoU: 31.4 -> 26.8 | Hurt |
| increase windscreen hazard from none to high | mapillary+ · WildDash windscreen subset | mIoU: 31.8 -> 27.8 | Hurt |
| increase particles hazard (fog, rain, snow) from none to high | mapillary+ · WildDash particles subset | mIoU: 30.2 -> 29.3 | No clear effect |
Closest papers by meaning
- SWE-fficiency: Can Language Models Optimize Real-World Repositories on Real Workloads? (ICML 2026 · similarity 0.75)
- RustMizan: A Compilable, Contamination-Aware Benchmarking Framework for Rust Vulnerabilities (NeurIPS 2026 · similarity 0.75)
- TestGenEval: A Real World Unit Test Generation and Test Completion Benchmark (ICLR 2025 · similarity 0.75)
- Precise Debugging Benchmark: Is Your Model Debugging or Regenerating? (NeurIPS 2026 · similarity 0.75)
- WildGuard: Open One-stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs (NeurIPS 2024 · similarity 0.74)
- Parallel Empirical Evaluations: Resilience despite Concurrency (AAAI 2024 · 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.
