A pre-segmented dataset for online human action detection and recognition from MoCap data, consisting of 15 subjects performing 15 exercises 2-3 times each. Capturing has been performed at 30 fps using one Microsoft Kinect sensor.

Download: CVD_dataset.zip