Accelerating Home Recovery: AI-Driven Accuracy Scoring of Physical-Therapy Exercises Using RGB Video from the KiMoRe Dataset

This project creates a model using RGB videos from the KiMoRe dataset (a dataset that includes five physician-selected lower-back exercises recorded from patients with Parkinson's, back pain, and a history of strokes) to generate clinical scores to show how well a patient performs an exercise per expert standards. By analyzing patient movements, the model produces scores across a range of rehabilitation activities, providing patients and physical therapists with rapid and accessible feedback to improve at-home recovery.

Interns: Rithika Kasireddy, Ahana Roy

mentor: Trystan May (AMDS/A3D)