GRF-MV:Ground Reaction Force Estimation from Monocular Video

Estimating ground reaction forces from monocular video has essential applications in healthcare, such as rehabilitation, injury prevention, patient monitoring, and physical therapy. However, it is a challenging problem due to the complexity of human motion, the limited availability of training data, and the difficulty of estimating contact and forces from only 2D monocular video. This paper presents a novel approach for estimating ground reaction forces (GRFs) from monocular video by combining deep learning-based 3D human mesh recovery with physics-based optimizations. Existing techniques for measuring GRFs rely on specialized sensors or multiple camera setups, limiting their applicability outside of lab settings. In contrast, our proposed approach requires only a single video camera, making it suitable for deployment in sports, clinical and home environments. A deep neural network is trained to recover 3D human mesh parameters from each frame, which are further refined using physics-based optimization (HybrIK- XL). GRFs are then estimated from the 3D foot velocities and contact modeling. The approach is evaluated on the GroundLink dataset, demonstrating improved accuracy over prior methods.