Submission: SurfEmb-PBR-RGBD/ITODD

Submission name
Submission time (UTC) Dec. 22, 2021, 6:09 p.m.
User surfemb
Dataset ITODD
Training model type Default
Training image type Synthetic (only PBR images provided for BOP Challenge 2020 were used)
Description
Evaluation scores
bop19_average_recall:0.538
bop19_average_recall_mspd:0.560
bop19_average_recall_mssd:0.558
bop19_average_recall_vsd:0.497
bop19_average_time_per_image:4.942

Method: SurfEmb-PBR-RGBD

User surfemb
Publication under review
Implementation will be made public soon
Training image modalities RGB
Test image modalities RGB-D
Description

SurfEmb: Dense and Continuous Correspondence Distributions for Object Pose Estimation with Learnt Surface Embeddings

Project site
Pre-print

Compared to the paper:

  • All models are trained for 500k iterations
  • 10k pose samples instead of 20k, and maximum 1k pose evaluations after pruning.

We use the available detections from CosyPose, but were not able to find detection-only timings. We add our timing on top of the full CosyPose single-view RGB pipeline, resulting in conservative timings.

Computer specifications 1 x RTX 2080