AI Paper Atlas

Research map / 3D vision / Human pose estimation

SGPA: Structure-Guided Prior Adaptation for Category-Level 6D Object Pose Estimation

Kai Chen, Qi Dou

ICCV 2021 · 3D vision › Human pose estimation · approach: Instance-level 6D object pose · 175 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add vanilla transformer prior adaptation (ablation)SGPA baseline · CAMERA25IoU75 mAP: 85.6 -> 87.8Improved
add vanilla transformer prior adaptation (ablation)SGPA baseline · CAMERA255°2cm mAP: 61.5 -> 68.2Improved
add vanilla transformer prior adaptation (ablation)SGPA baseline · REAL275IoU75 mAP: 59.2 -> 59.9Improved
add vanilla transformer prior adaptation (ablation)SGPA baseline · REAL2755°2cm mAP: 23.8 -> 33.3Improved
replace vanilla transformer with structure regularized low-rank transformer (ablation)SGPA · REAL2755°2cm mAP: 33.3 -> 35.9Improved
replace vanilla transformer with structure regularized low-rank transformer (ablation)SGPA · CAMERA255°2cm mAP: 68.2 -> 70.7Improved
use random instance as prior instead of learned embedding prior (ablation)SGPA · REAL2755°2cm mAP: 35.9 -> 33.9Hurt
use random instance as prior instead of learned embedding prior (ablation)SGPA · CAMERA255°2cm mAP: 70.7 -> 69.5No clear effect
reduce low-rank dimension n from 256 to 16 (ablation)SGPA · REAL2755°2cm mAP: 35.9 -> 27.7Hurt
use full rank 1024 instead of n=256 (ablation)SGPA · REAL2755°2cm mAP: 35.9 -> 33.3Hurt
vary low-rank dimension n from 16 to 1024 (ablation)SGPA · CAMERA255°2cm mAP: 68.2 -> 70.7No clear effect
use SGPA prior adaptation instead of SPDSPD · REAL2755°2cm mAP: 19.3 -> 35.9Improved

Later papers that compared with this method

9 later accepted papers report SGPA as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.

1 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy