Evaluation dataset staged in a home kitchen to benchmark how well the RT-DETR model detects and identifies vegetables inside a densely packed refrigerator. 10 renders at 640x640 with albedo and metric depth passes plus per-frame annotations (bounding boxes, labels, camera pose).
This dataset mirrors public data-pack render outputs from Physicl.
Each row represents one render view. The image column contains a stable URL to the… See the full description on the dataset page:
https://huggingface.co/datasets/physicl-test/pack-8db-of-my-project-r2a9xm-34f47df3.
