Submission: NOCS_prune/XYZ-IBD/Synth train (default bboxs)

Download submission
Submission name Synth train (default bboxs)
Submission time (UTC) May 29, 2026, 3:04 p.m.
User GRAA
Task Model-based 6D detection of seen objects
Dataset XYZ-IBD
Description - trained on given synthetic data - default bbox used
Evaluation scores
AP:0.857
AP_25:0.867
AP_25_mm:0.768
AP_MSPD:0.847
AP_MSSD:0.867
AP_MSSD_mm:0.768
average_time_per_image:-1.000

Method: NOCS_prune

User GRAA
Publication
Implementation
Training image modalities RGB-D
Test image modalities RGB-D
Description

NOCS coordinates were predicted from 2D image, and 3D–3D correspondences were established. Noisy points and inaccurate predictions were removed, and the Kabsch algorithm was used for 3D pose estimation.

Computer specifications 1080 Ti