AI Paper Atlas

Research map / 3D vision / Human pose estimation

Learning Causal Representation for Training Cross-Domain Pose Estimator via Generative Interventions

Xiheng Zhang, Yongkang Wong, Xiaofei Wu, Juwei Lu, Mohan Kankanhalli, Xiangdong Li, Weidong Geng

ICCV 2021 · 3D vision › Human pose estimation · approach: Monocular 2D-to-3D pose lifting · 30 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
add counterfactual feature generator with unlabeled target data (SD + TD)pose estimator HRNet-W32 · H3.6M->3DPWMPJPE: 118.7 -> 94.7Improved
use MMD instead of naive l2 as discrepancy loss (ablation)SD + TD model · H3.6M->3DPWMPJPE: 100.3 -> 94.7Improved
use MMD instead of KL divergence as discrepancy loss (ablation)SD + TD model · H3.6M->3DPWMPJPE: 96.5 -> 94.7Improved
apply method on ResNet-18 backboneSD + TD model · H3.6M->3DPWMPJPE: 122.4 -> 98.3Improved
apply method on ResNet-50 backboneSD + TD model · H3.6M->3DPWMPJPE: 120.2 -> 96.3Improved
add 3DHP, HumanEva and SURREAL as extra source datasets (ablation)SD + TD model · H3.6M->3DPWMPJPE: 94.7 -> 94.1No clear effect
add H3.6M, HumanEva and SURREAL as extra source datasets (ablation)SD + TD model · 3DPW->3DHPMPJPE: 109.7 -> 102.7Improved
use multiple unconstrained datasets (UCF101, HMDB, Kinetics) instead of Kinetics only (ablation)SD + UD model · H3.6M->3DPWMPJPE: 97.5 -> 94.9Improved
use all three general image datasets instead of ImageNet only as interventions (ablation)SD + UD model · H3.6M->3DPWMPJPE: 102.3 -> 99.2Improved
use unconstrained data (Kinetics) instead of target datadomain generalization · H3.6M->3DPWMPJPE: 94.7 -> 97.5Hurt
add counterfactual features with unlabeled target data on hand poseFreiHAND model · FreiHAND->STBEPE: 36.1 -> 22.4Improved
single unconstrained intervention dataset (UCF101 vs ImageNet) (ablation)SD + UD model · H3.6M->3DPWMPJPE: 98.6 vs 102.3Mixed

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