AI Paper Atlas

Research map / 3D vision / Human pose estimation

NeMo: Neural Mesh Models of Contrastive Features for Robust 3D Pose Estimation

Angtian Wang, Adam Kortylewski, Alan Yuille

ICLR 2021 · 3D vision › Human pose estimation · approach: Category-level object pose estimation · 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
Use feature-level NeMo rendering with occlusion/outlier inference instead of StarMap3D pose estimation · PASCAL3D+ L3ACC π/6 (%): 22.9 -> 47.1Improved
Use feature-level NeMo rendering with occlusion/outlier inference instead of StarMap3D pose estimation · PASCAL3D+ L2ACC π/6 (%): 47.2 -> 65.2Improved
Use feature-level NeMo rendering with occlusion/outlier inference instead of StarMap3D pose estimation · PASCAL3D+ L0ACC π/6 (%): 89.4 -> 86.7Hurt
Replace detailed mesh with category-specific cuboid meshesNeMo · PASCAL3D+ L0ACC π/6 (%): 84.1 -> 86.7Improved
Replace detailed mesh with one category-level cuboidNeMo · PASCAL3D+ L2ACC π/6 (%): 59.9 -> 63.9Improved
Remove outlier/clutter handling during inference (ablation)NeMo-MultiCuboid · PASCAL3D+ L3ACC π/6 (%): 47.1 -> 44.4Hurt
Add outlier/clutter handling during inference (ablation)NeMo-MultiCuboid · PASCAL3D+ L3Median error (degree): 41.6 -> 36.1Improved
Remove contrastive feature loss (ablation)NeMo-MultiCuboid · PASCAL3D+ L3ACC π/6 (%): 47.1 -> 26.9Hurt
Add contrastive feature loss (ablation)NeMo-MultiCuboid · PASCAL3D+ L3ACC π/18 (%): 5.6 -> 17.8Improved
Reduce random pose initialization samples from 144 to 72 (ablation)NeMo-MultiCuboid · PASCAL3D+ACC π/6 (%): 86.7 -> 86.3No clear effect
Reduce random pose initialization samples from 144 to 1 (ablation)NeMo-MultiCuboid · PASCAL3D+ACC π/6 (%): 86.7 -> 54.9Hurt
Use category-specific subtype information at inference (ablation)NeMo-MultiCuboid · PASCAL3D+ L3ACC π/6 (%): 47.2 -> 47.1No clear effect

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