Research map / 3D vision / Human pose estimation
AIFit: Automatic 3D Human-Interpretable Feedback Models for Fitness Training
Mihai Fieraru, Mihai Zanfir, Silviu Cristian Pirlea, Vlad Olaru, Cristian Sminchisescu
Check 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 |
|---|---|---|---|
| Fine-tune MubyNet on Fit3D | MubyNet 3D pose reconstruction · Fit3D test | MPJPE (mm), w/o Procrustes, SV: 90.4 -> 52.4 | Improved |
| Use multi-view instead of single-view reconstruction | MubyNet · Fit3D test | MPJPE (mm), w/o Procrustes: 90.4 -> 71.9; 52.4 -> 45.4 | Improved |
| Fine-tune MubyNet on Fit3D | MubyNet 3D pose reconstruction · Fit3D test | MPJPE (mm), w. Procrustes, SV: 67.7 -> 41.1 | Improved |
| Use multi-view instead of single-view reconstruction | MubyNet · Fit3D test | MPJPE (mm), w. Procrustes: 67.7 -> 57.9; 41.1 -> 35.7 | Improved |
| Use ground-truth 3D joints rather than estimated poses | AIFit repetition segmentation · Fit3D test | Repetition segmentation accuracy (IoU): 0.730 -> 0.731 | Improved |
| Use fine-tuned multi-view pose estimates instead of single-view MubyNet estimates | AIFit repetition segmentation · Fit3D test | Repetition segmentation accuracy (IoU): 0.661 -> 0.730 | Improved |
| Use 3D pose representation rather than 2D pose representation (ablation) | AIFit repetition counting · Fit3D | OBO error: 0.920 -> 0.140 | Improved |
| Use AIFit repetition counting instead of RepNet | Repetition counting · Fit3D test | OBO error: 0.520 -> 0.140 | Improved |
| Use AIFit repetition counting instead of RepNet | Repetition counting · Fit3D test | MAE: 0.740 -> 0.253 | Improved |
| Use AIFit repetition counting instead of RepNet | Repetition counting · CountixFitness validation | OBO error: 0.292 -> 0.292 | No clear effect |
| Use AIFit repetition counting instead of RepNet | Repetition counting · CountixFitness validation | MAE: 0.468 -> 0.604 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- SPIN — SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition (AAAI 2021) · better on 4 of 4 shared results
- RepNet — RepNet: Weakly Supervised Training of an Adversarial Reprojection Network for 3D Human Pose Estimation (CVPR 2019) · better on 2 of 4 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
- Domain Knowledge-Informed Self-Supervised Representations for Workout Form Assessment (ECCV 2022 · 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 0.89)
- 3D Human Pose Estimation with Muscles (NeurIPS 2025 · similarity 0.88)
- FLAG3D: A 3D Fitness Activity Dataset With Language Instruction (CVPR 2023 · similarity 0.87)
- ExpertAF: Expert Actionable Feedback from Video (CVPR 2025 · similarity 0.87)
- From 3D Pose to Prose: Biomechanics-Grounded Vision-Language Coaching (CVPR 2026 · similarity 0.86)
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.
