AI Paper Atlas

Research map / 3D vision / Human pose estimation

MuCHeR: Multi-Person Camera-Centric Human Detection, Mesh Recovery and Tracking

Guénolé Fiche, Philippe Weinzaepfel, Romain Brégier, Fabien Baradel

ECCV 2026 · 3D vision › Human pose estimation · approach: Human mesh recovery from images

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
match queries on 2D location instead of 3D location (ablation)Multi-HMR 2 · 3DPWPVE: 93.0 -> 90.5No clear effect
match queries on 2D location instead of 3D joints (ablation)Multi-HMR 2 · 3DPWPVE: 104.5 -> 90.5Improved
2D-location matching (only matched queries go through body model) (ablation)Multi-HMR 2 · training costforward-backward time: 0.48s -> 0.35sImproved
2D-location matching (ablation)Multi-HMR 2 · training costGPU memory: 27GB -> 7GBImproved
add 2D pseudo-labeled data to synthetic+real training (ablation)Multi-HMR 2 · Hi4DMPJPE: 60.0 -> 76.5Hurt
add 2D pseudo-labeled data to synthetic+real training (ablation)Multi-HMR 2 · MSCOCOAR: 29.4 -> 32.9Improved
add angular 2D loss (ablation)Multi-HMR 2 · Hi4DMPJPE: 70.9 -> 60.0Improved
add angular 2D loss (ablation)Multi-HMR 2 · MSCOCOAP: 23.7 -> 21.9Hurt
train with diverse aspect ratios instead of square only (ablation)Multi-HMR 2 · MSCOCO (test w/o padding)AP: 0.5 -> 21.9Improved
test with predicted instead of GT camera (GT-trained) (ablation)Multi-HMR 2 · 3DPWAbs-PVE: 243.7 -> 609.8Hurt
train and test with predicted camera instead of GT-train/pred-test (ablation)Multi-HMR 2 · Hi4DPair-PA-MPJPE: 91.0 -> 72.7Improved
increase decoder blocks from 1 to 8 (ablation)Multi-HMR 2 · 3DPWPVE: 96.4 -> 90.5Improved

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