Benchmark RadarCOD10K-C
COD10K-C evaluates camouflaged object detection models under 8 corruption types at 5 severity levels, yielding 40 conditions and 81,040 image pairs based on COD10K. Models are scored on Dice and other standard metrics for segmentation robustness.
- Released
- 2026-05-23
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Standard camouflaged object detection benchmarks measure performance on clean images only, while real-world captures include blur, noise, weather, and compression artifacts. This benchmark quantifies robustness drops under such corruptions, enabling selection of models that degrade gracefully in practical conditions.
Motivation
Camouflaged object detection has improved substantially, but most standard benchmarks evaluate models only on clean images.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.