AI Paper Atlas

Research map / 3D vision / Human pose estimation

Glissando-Net: Deep Single View Category Level Pose Estimation and 3D Reconstruction

Bo Sun, Hao Kang, Li Guan, Haoxiang Li, Philippos Mordohai, Gang Hua

TPAMI 2024 · 3D vision › Human pose estimation · approach: Category-level object pose estimation

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 the 3D point cloud encoder (train with ground-truth shape) (ablation)Glissando-Net (EMD) · NOCS real overallmean Chamfer distance (mm): 2.62 -> 0.85Improved
add the 3D point cloud encoder (ablation)Glissando-Net (EMD) · NOCS real overall10deg & 10cm accuracy: 2.97% -> 19.08%Improved
use EMD instead of Chamfer distance as shape loss (ablation)Glissando-Net · NOCS real overallmean Chamfer distance (mm): 0.91 -> 0.85Improved
use EMD instead of Chamfer distance as shape loss (ablation)Glissando-Net · NOCS real overall10deg & 10cm accuracy: 16.27% -> 19.08%Improved
increase latent code size from 32 to 256 (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)Chamfer (mm): 0.98 -> 0.81Improved
increase latent code size from 32 to 256 (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)10deg & 10cm accuracy: 6.67% -> 18.50%Improved
increase latent code size beyond 256 (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)Chamfer (mm): 0.81 -> 0.80No clear effect
add a pose regression head after the shape decoder (vs none) (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)Chamfer (mm): 0.83 -> 0.81Improved
place pose head after shape decoder rather than image encoder (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)10deg & 10cm accuracy: 15.67% -> 18.50%Improved
use separate networks for rotation and translation (ablation)Glissando-Net · NOCS synthetic ablation (10K subset)10deg & 10cm accuracy: 10.20% -> 18.50%Improved
add feature transform between RGB decoder and point cloud encoder (ablation)Glissando-Net (EMD) · NOCS syntheticChamfer: 0.94 -> 0.62Improved
add feature transform between RGB decoder and point cloud decoder (ablation)Glissando-Net (EMD) · NOCS synthetic10deg & 10cm accuracy: 2.06% -> 31.86%Improved

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