mlsanityData to 14 Oct 2026

Cross-Model Distillation of a Human-Pose Foundation Model from Unannotated Infant Video for Markerless 3D Pose Estimation

arXiv 2609.01840 ↗First posted 1 Sep 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Spontaneous movement is one of the earliest windows onto an infant's neuromotor health, and structured clinical instruments that score it are validated early predictors of cerebral-palsy risk. However, they require specially trained raters, are time-consuming, and carry inter-rater variability. This motivates automated, video-based markerless assessment, especially as marker-based motion capture is impractical in infants. Yet the foundation models that make markerless capture possible are trained almost entirely on adults: our recent multi-view infant study found that no single model is jointly best, with strong 2D keypoint accuracy and direct 3D body recovery split across different models. While that study identifies this trade-off, it does not resolve it. Here, we perform cross-model distillation from the Sapiens 2 pose model into the SAM 3D Body model, using unannotated infant video alone. A frozen teacher supplies dense pseudo-labels, and a differentiable renderer aligns the predicted mesh to them in the training loop. On eleven held-out infants (18 sessions, 173 recordings) under our prior study's multi-view protocol, fine-tuning improves same-view 2D keypoint agreement with the Sapiens reference (median body percentage of correct keypoints @ 10px 0.22 -> 0.42, face 0.22 -> 0.42) and Procrustes-aligned mean per joint 3D position error (25.5 -> 22.2 mm). This demonstrates how cross-model distillation improves SAM 3D Body model performance on infants.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 18 of its 35 references that are in the map. First posted on arXiv 1 Sep 2026.

Builds on

18 of 35 references.

References · 18

  1. 2019Expressive Body Capture: 3D Hands, Face, and Body from a Single Imagecited by 1,612
  2. 2025SAM 3: Segment Anything with Conceptscited by 906
  3. 2024Sapiens: Foundation for Human Vision Modelscited by 229
  4. 2021Equinox: neural networks in JAX via callable PyTrees and filtered transformationscited by 148
  5. 2026SAM 3D Body: Robust Full-Body Human Mesh Recoverycited by 69
  6. 2022Learning 3D Human Pose Estimation from Dozens of Datasets using a Geometry-Aware Autoencoder to Bridge Between Skeleton Formatscited by 50
  7. 2020Invariant Representation Learning for Infant Pose Estimation with Small Datacited by 20
  8. 2018Learning and Tracking the 3D Body Shape of Freely Moving Infants from RGB-D sequencescited by 19
  9. 2024Differentiable Biomechanics Unlocks Opportunities for Markerless Motion Capturecited by 18
  10. 2023Improved Trajectory Reconstruction for Markerless Pose Estimationcited by 12
  11. 2022On Triangulation as a Form of Self-Supervision for 3D Human Pose Estimationcited by 10
  12. 2025Human Mesh Modeling for Anny Bodycited by 10
6 more
  1. 2024Automatic infant 2D pose estimation from videos: comparing seven deep neural network methodscited by 6
  2. 2026Monocular Biomechanical Tracking of Fingers with Inverse Kinematics to Foundation Modelscited by 5
  3. 2026Sapiens2cited by 4
  4. 2025Leveraging Synthetic Adult Datasets for Unsupervised Infant Pose Estimationcited by 2
  5. 2026Markerless Motion Capture for Biomechanical Whole-Body Kinematic Estimation in Infantscited by 1
  6. 2026Biomechanical 3D Body: Self-Supervised Distillation of Biomechanical Pose from a 3D Body Foundation Modelcited by 1

Built on by

Not cited yet.