Pediatric locomotion analysis holds potential for early detection of neurodevelopmental disorders; however, current methods for data collection in toddlers are costly and challenging. We present a computational pipeline for extracting gait metrics from standard 2D videos. The pipeline integrates state-of-the-art pose estimation models, world-coordinate alignment via WHAM, and gait metric validation against Motion Capture ground truth. Applied to videos of toddlers walking, we will evaluate whether early gait markers detected by the pipeline correlate with later autism spectrum disorder diagnoses. This approach enables motion analysis without specialized equipment, expanding pediatric motor control research and supporting earlier intervention for disorders with motor manifestations.

I have worked in the Seethapathi lab for two years, gaining experience in computational methods for studying motor control and its clinical translation. Through SuperUROP, I aim to secure dedicated time and mentorship to explore the diagnostic potential of my computer vision pipeline. I hope to enhance my project leadership skills, refine my research practices, and develop a tangible product while collaborating closely with my lab.