Radroachhd. Jun 2026

We define a taxonomy of 12 classes specific to hazard navigation:

We propose , a dataset and benchmark specifically curated to simulate the visual challenges of a "wasteland" environment. While the name is derived from colloquial terminology for radiological pests (the "radroach"), the dataset encompasses a broader scope: it focuses on High Definition (HD) capture of environmental hazards, obscured pathways, and object detection in the presence of visual noise (dust, smoke, radiation haze). radroachhd.