Research map / 3D vision / Human pose estimation
Domain Knowledge-Informed Self-Supervised Representations for Workout Form Assessment
Paritosh Parmar, Amol Gharat, Helge Rhodin
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use cross-view cross-subject triplets instead of single-view single-subject (Vanilla PC) (ablation) | pose contrastive learning · Fitness-AQA BackSquat KIE | F-score: 0.4118 -> 0.5195 | Improved |
| use cross-view cross-subject triplets instead of single-view single-subject (Vanilla PC) (ablation) | pose contrastive learning · Fitness-AQA BackSquat KFE | F-score: 0.7965 -> 0.8286 | Improved |
| self-supervised pose contrastive training on gym videos instead of plain ImageNet features | ResNet-18 · Fitness-AQA BackSquat KIE | F-score: 0.1923 -> 0.5195 | Improved |
| domain-informed motion disentangling instead of Kinetics-pretrained features | R(2+1)D-18 · Fitness-AQA BackSquat KIE | F-score: 0.2970 -> 0.4186 | Improved |
| contrastive pose learning instead of reconstruction-based PAD | image model · Fitness-AQA BackSquat KIE | F-score: 0.3180 -> 0.5195 | Improved |
| replace OpenPose-TDM pose features with CVCSPC | OpenPose-TDM · Fitness-AQA BackSquat KIE | F-score: 0.4143 -> 0.5195 | Improved |
| replace OpenPose-TDM with CVCSPC | OpenPose-TDM · Fitness-AQA BackSquat KFE | F-score: 0.8123 -> 0.8286 | No clear effect |
| ensemble CVCSPC and MD (ablation) | CVCSPC · Fitness-AQA BackSquat KIE | F-score: 0.5195 -> 0.5263 | Improved |
| replace HMR-TDM pose features with CVCSPC | HMR-TDM · Waseda Squat (average) | accuracy (%): 89.08 -> 89.92 | No clear effect |
| replace HMR-TDM with CVCSPC on shallow squat | HMR-TDM · Waseda Squat (SS) | accuracy (%): 87.30 -> 77.77 | Hurt |
| replace OpenPose-TDM with CVCSPC | OpenPose-TDM · Fitness-AQA BackSquat shallow squat | F-score: 0.8340 -> 0.8694 | Improved |
| replace OpenPose-TDM with MD on knees error | OpenPose-TDM · Fitness-AQA OverheadPress | F-score: 0.7131 -> 0.8452 | Improved |
Closest papers by meaning
- AIFit: Automatic 3D Human-Interpretable Feedback Models for Fitness Training (CVPR 2021 · similarity 0.89)
- M3GYM: A Large-Scale Multimodal Multi-view Multi-person Pose Dataset for Fitness Activity Understanding in Real-world Settings (CVPR 2025 · similarity 0.88)
- ExpertAF: Expert Actionable Feedback from Video (CVPR 2025 · similarity 0.85)
- From 3D Pose to Prose: Biomechanics-Grounded Vision-Language Coaching (CVPR 2026 · similarity 0.84)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.84)
- Learning Skill-Attributes for Transferable Assessment in Video (NeurIPS 2025 · similarity 0.84)
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.
