AI Paper Atlas

Research map / 3D vision / Human pose estimation

Glimpse: Geometry Learning of Multi-scale Structural Priors for 3D Pose Estimation

Zhenhua Tang, Jihua Peng, Yanbin Hao, Qiguang Miao, Chi-Man Pun

ICML 2026 · 3D vision › Human pose estimation · approach: Monocular 2D-to-3D pose lifting

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 joint-level deformable sampling of image features (ablation)Glimpse-HR32 baseline · Human3.6MMPJPE: 51.2 -> 41.4Improved
add limb-level sampling (M=3) (ablation)Glimpse-HR32 with joint sampling · Human3.6MMPJPE: 41.4 -> 40.3Improved
add geometric correction to joint sampling (ablation)Glimpse-HR32 · Human3.6MMPJPE: 41.4 -> 40.3Improved
combine joint sampling, limb sampling and geometric correction (ablation)Glimpse-HR32 · Human3.6MMPJPE: 40.3 -> 39.5Improved
set limb points M to 3 versus 1 (ablation)Glimpse-HR32 · Human3.6MMPJPE: 40.6 -> 40.3Improved
set limb points M to 3 versus 4 (ablation)Glimpse-HR32 · Human3.6MMPJPE: 40.4 -> 40.3Improved
replace CA-PF-HR32 with Glimpse-HR32CA-PF-HR32 · Human3.6MMPJPE: 41.4 -> 39.5Improved
replace CA-PF-HR32 with Glimpse-HR32 under occlusionCA-PF-HR32 · Human3.6M Sitting Down S9MPJPE: 65.5 -> 53.1Improved
replace CA-PF-HR32 with Glimpse-HR32 under noise sigma=5.0CA-PF-HR32 · Human3.6M noisy 2D inputMPJPE: 192.9 -> 104.9Improved
replace CA-PF-HR32 with Glimpse-HR32 cross-datasetCA-PF-HR32 · 3DPWMPJPE: 138.9 -> 132.2Improved
replace CA-PF-HR48 with Glimpse-HR48CA-PF-HR48 · MPI-INF-3DHPAUC: 76.3 -> 80.9Improved
use HRNet-W48 instead of W32 backboneGlimpse · Human3.6MMPJPE: 39.5 -> 39.3Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

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