TUM-GAID, short for TUM Gait from Audio, Image and Depth, is a multimodal gait dataset. The dataset includes two walking trajectories of 305 subjects in an indoor environment. The first trajectory is from left to right, and the second trajectory is from right to left.

Two recording sessions were conducted, one in January when the subjects wore thick jackets and most wore winter boots, and another in April when the subjects wore lighter clothing.

The motion is captured by a Microsoft Kinect sensor, which provides a video stream with a resolution of 640 × 480 pixels and a frame rate of approximately 30 FPS.