Model-based 2D detection of unseen objects – BOP-H3

This leaderbord shows the overall ranking for Model-based 2D detection of unseen objects on BOP-H3 (HOT3D, HOPEv2, HANDAL). For each method, the date of the latest considered submission is reported. If more submissions of a method are available for a dataset D, the submission with the highest APD score is considered. The reported time is the average time to generate predictions for all objects in a single image (averaged over the datasets).

Date (UTC) Method Test image APH3 APHOT3D APHOPEv2 APHANDAL Time (s)
2025-08-26 MUSE RGB 0.418 0.438 0.460 0.357 0.919
2024-09-23 CNOS (FastSAM) RGB 0.303 0.350 0.313 0.246 0.332
2025-10-01 3PT-Detection (f.k.a IPT) RGB 0.542 0.481 0.618 0.527 58.249
2025-11-14 3PT-Detection (Nov 2025) RGB 0.548 0.484 0.624 0.536 6.551
2024-09-23 CNOS (SAM) RGB 0.317 0.197

Tip: Hover over the numbers to see more decimal places.

Latest submissions to this leaderboard: