Submission: CosyPose-ECCV20-SYNT+REAL-1VIEW-ICP/HB

Download submission
Submission name
Submission time (UTC) Aug. 17, 2020, 11:11 p.m.
User yann_labbe
Task 6D localization of seen objects
Dataset HB
Training model type Default
Training image type Synthetic (only PBR images provided for BOP Challenge 2020 were used)
Description
Evaluation scores
AR:0.712
AR_MSPD:0.737
AR_MSSD:0.719
AR_VSD:0.679
average_time_per_image:5.326

Method: CosyPose-ECCV20-SYNT+REAL-1VIEW-ICP

User yann_labbe
Publication Labbé et al, CosyPose: Consistent multi-view multi-object 6D pose estimation, ECCV 2020
Implementation https://github.com/ylabbe/cosypose
Training image modalities RGB
Test image modalities RGB-D
Description

We apply ICP to each individual 6D pose estimates. The pose estimates are the ones from CosyPose-ECCV20-SYNT+REAL-1VIEW. We use parts of Pix2pose's ICP implementation https://github.com/kirumang/Pix2Pose/blob/master/tools/5_evaluation_bop_icp3d.py.

MaskRCNN predicted masks are used to pick the 3D points that correspond to the detection in the depth.

Computer specifications CPU: 20-core Intel Xeon 6164 @ 3.2 GHz, GPU: Nvidia V100