| Abstract | The Multi-View Hand gesture (MuViH) dataset comprises RGB-D images and annotations of participants performing various hand gestures in an industrial-type environment. Our aim is to provide a set of images tailored to the field of scene supervision and gesture recognition applied to interactive manufacturing processes using robotics. Compared to existing publicly available datasets, the MuViH dataset provides greater variability in the visual content: changing background according to camera viewpoint, variable viewing angle and distance to the scene, morphology of participants, occlusions caused by the robot or other equipment, participant's position, orientation and pose within the field of view. The data collection took place during the summer of 2023, at the Centre de technologies de fabrication en aérospatiale (CTFA) located in Montreal, Canada. In all, 20 individuals participated in this activity. All the acquisitions took place in a cobotic cell, i.e. an enclosed area containing an industrial robot. The file 'Overview_MuViH.md' provides more details on the dataset and how to access it. Also provided here are some sample images from the dataset and an illustration of the gesture dictionary. |
|---|
