AI Paper Atlas

Research map / 3D vision / Human pose estimation

Omni6D: Large-Vocabulary 3D Object Dataset for Category-Level 6D Object Pose Estimation

Mengchen Zhang, Tong Wu, Tai Wang, Tengfei Wang, Ziwei Liu, Dahua Lin

ECCV 2024 · 3D vision › Human pose estimation · approach: Instance-level 6D object pose · 13 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
fine-tune from a CAMERA-pretrained model with progressive category expansion instead of training from scratch (ablation)SPD, DualPoseNet, HS-Pose · Omni6D cls3 to cls485deg2cm, 5deg, 2cmImproved
increase the number of training categories from 3 to 48 (ablation)HS-Pose (fine-tuned) · Omni6D cls3 to cls485deg2cm: 62.52% -> 14.42%Hurt
add Omni6D data to REAL data for training (ablation)DualPoseNet · REAL2755deg2cm: 8.76 -> 14.10Improved
train on Omni6D only and test on real data (ablation)DualPoseNet · REAL2755deg2cm: 6.55 vs 8.76 (REAL only)Mixed
use symmetry-aware metric instead of non-symmetry-aware metric (ablation)Omni6D evaluation · Omni6D testpose accuracyImproved
implicit shape-prior methods tested on unseen categoriesSPD, SGPA · Omni6D-outIoU50: SPD 44.56 -> 7.56Hurt
higher instance diversity within a category (mean Chamfer distance)category-level pose methods · Omni6D per-category analysisIoU75 and 5deg2cmImproved
more instances in a shape cluster rather than a semantic categorycategory-level pose methods · Omni6D per-category analysisIoU75 and 5deg2cmImproved

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