Replication Data for: Deep learning-based physical exercise assessment of older adults using single-camera videos

This dataset contains multimodal recordings of older adults performing structured physical exercises in a residential care setting, collected to support research in motion analysis and exercise performance monitoring. It includes synchronized 2D pose data extracted from video, joint angles with 3 rotational degrees of freedom derived from IMU-based motion capture, and detailed sub-exercise annotations, enabling the development and evaluation of machine learning models for exercise classification and performance assessment. The dataset focuses on six functional exercises relevant to physical health monitoring and provides fine-grained action segmentation to facilitate longitudinal and unsupervised video assessment of exercise performance in older populations.