EgoGesture is a dataset for large-scale gesture recognition. The dataset is large, varied, and realistic enough to train deep neural networks. The dataset is used for egocentric gesture recognition. The dataset provides a testbed not only for gesture classification in segmented data, but also for gesture detection in continuous data.

The dataset includes:

• 2,081 RGB-D videos
• 2,4161 gesture samples
• 2,953,224 frames from 50 different subjects