Here are some pictures of a robot costume wearing high-specularity edge-case mirror suit, a dataset (425 RAW/JPEGs) for benchmarking CV & depth-estimation algorithms against extreme mirror reflections [D]

A new dataset, comprising 425 RAW/JPEG assets, has been released to stress-test computer vision models, depth cameras, and spatial AI. This dataset features a custom faceted mirror suit, captured in high-contrast outdoor environments, designed to induce bounding-box dropouts and segmentation failures. It includes uncompressed Camera-Master RAWs, high-resolution JPEGs, and SHA-256 forensic manifests, providing a robust resource for benchmarking algorithms against extreme mirror reflections and specular glare.

Unlike previous datasets, this one is purpose-built with a custom faceted mirror suit to specifically stress-test computer vision models against severe specular glare and geometric reflections.