Intel RealSense Activity Dataset (IRAD)

We introduce the Intel RealSense Multimodal Activity Dataset (IRMAD), a custom collection designed to evaluate multimodal human activity recognition methods under realistic conditions. IRMAD comprises synchronous captures of four data streams, like 2D RGB, 2D depth, 3D RGB point clouds, and 3D depth point clouds, acquired with an Intel RealSense D435 camera. The dataset has 7 activity classes, like `sitting', `standing', `walking', `running', `boxing', `hand-waving', and `falling down'. The dataset comprises data from 15 individuals. This includes 10 male and 5 female participants, with ages ranging between 23 and 34 years.