AI Paper Atlas

Research map / 3D vision / Human pose estimation

FreiHAND: A Dataset for Markerless Capture of Hand Pose and Shape From Single RGB Images

Christian Zimmermann, Duygu Ceylan, Jimei Yang, Bryan Russell, Max Argus, Thomas Brox

ICCV 2019 · 3D vision › Human pose estimation · approach: Depth-based hand pose regression · 480 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 background compositing (Cut&Paste) to green screen images (ablation)FreiHAND training · FreiHAND eval (from scratch)3D PCK AUC: 0.440 -> 0.508Improved
add harmonization to Cut&Paste (ablation)FreiHAND training · FreiHAND eval (from scratch)3D PCK AUC: 0.508 -> 0.628Improved
combine all post-processing options jointly (ablation)FreiHAND training · FreiHAND eval (from scratch)3D PCK AUC: 0.643 -> 0.678Improved
add compositing post-processing with a pretrained ResNet50 (ablation)FreiHAND training · FreiHAND eval (ResNet50 pretrained)3D PCK AUC: 0.436 -> 0.523Improved
add compositing post-processing when training on RHD (ablation)RHD training · FreiHAND eval3D PCK AUC: 0.508 -> 0.514No clear effect
add post-processing to RHD training, evaluated on RHD (ablation)RHD training · RHD eval3D PCK AUC: 0.767 -> 0.756Hurt
grow the labeled dataset from D0 to D4 through iterative labeling (ablation)FreiHAND · RHD eval3D PCK AUC: 0.244 -> 0.518Improved
grow the labeled dataset from D0 to D4 through iterative labeling (ablation)FreiHAND · PAN eval3D PCK AUC: 0.347 -> 0.562Improved
predict MANO parameters with a CNN instead of fitting MANO to predicted keypoints (ablation)shape estimation · FreiHAND evalmesh error: 1.44 -> 1.07Improved
predict MANO parameters with a CNN instead of fitting MANO to predicted keypoints (ablation)shape estimation · FreiHAND evalF@5mm: 0.416 -> 0.529Improved
use ImageNet-pretrained ResNet50 instead of from-scratch network, trained on FreiHANDFreiHAND training · FreiHAND eval3D PCK AUC: 0.678 -> 0.523Hurt
use Scorevolume loss instead of Softargmax loss for MVNet (ablation)MVNet · labeling confidencesparsification curveMixed

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