Deep learning and computer vision for 3D human understanding, medical imaging, and real-world deployment.
We benchmark three state-of-the-art 3D pose estimation models on biomechanical joint angle accuracy and discover that models with lower MPJPE produce higher joint angle error — challenging a core assumption in the field.
A large-scale 3D pose estimation dataset of 5,966 professional tennis serves from broadcast video, enabling motion classification, player identification, and performance prediction using deep learning at unprecedented scale.
Benchmarking state-of-the-art pose estimation algorithms across camera configurations for real-world 3D human pose analysis.
A deep learning approach for predicting pediatric Crohn's disease from histopathological whole slide images, with investigation of inter-site generalizability and site-specific artifact mitigation.