Understanding infant head and neck motion is important for biomechanical research on abusive head trauma, including shaken baby syndrome. Finite element models can help investigate injury mechanisms, but their predictions depend on the motion prescribed as boundary conditions. Video-based motion reconstruction offers a potential source of such information. However, developing and evaluating these methods is challenging because real infant motion data with accurate three-dimensional ground truth are scarce, and potentially harmful loading conditions cannot be reproduced experimentally with infants. This project aims to develop a biomechanically informed synthetic dataset for infant pose estimation and head motion reconstruction. The proposed workflow combines infant and adult body models, physics-based simulation, and image rendering to generate annotated sequences of adult–infant interactions, including controlled shaking and fall scenarios. Infant head–neck mechanical parameters will be calibrated against reference finite element simulations. Rendering will introduce variations in camera viewpoints, lighting, appearance, and background to represent different visual conditions. The generated sequences will provide images together with corresponding three-dimensional poses and head motion trajectories. These data will support the training and evaluation of pose estimation methods under controlled conditions. Experiments will examine how viewpoint, occlusion, and visual variation affect reconstruction accuracy, and how pose estimation errors propagate into derived velocities and accelerations. The requested computational resources will support simulation, synthetic data generation, and model training and evaluation. The expected outcome is an annotated dataset and a reproducible workflow for assessing video-based infant motion reconstruction, providing a foundation for future validation and the development of motion inputs for biomechanical injury analysis. Main supervisor: Xiaogai Li, CBH, KTH Royal Institute of Technology.