ExoNet: Egocentric Images of Walking Environments

Computer vision can be used for environment-adaptive control of robotic exoskeletons and prostheses. However, small-scale and private training datasets have impeded the development of image classification algorithms (e.g., convolutional neural networks) to recognize the walking environment. To address these limitations, we developed ExoNet, a large-scale dataset of wearable camera images (i.e., egocentric perception) of real-world walking environments. ExoNet contains over 5.6 million RGB images of indoor and outdoor walking environments, which were collected throughout the summer, fall, and winter seasons. Approximately 923,000 images in ExoNet were human-annotated using a 12-class hierarchical labelling architecture. Available publicly through IEEE DataPort, ExoNet serves as a communal platform to train, develop, and compare image classification algorithms for visual perception of human walking environments. In addition to robotic exoskeletons and prostheses, applications of ExoNet can extend to humanoids and autonomous legged robots. 

Reference:

Laschowski B, McNally W, Wong A, and McPhee J. (2022). Environment Classification for Robotic Leg Prostheses and Exoskeletons using Deep Convolutional Neural Networks. Frontiers in Neurorobotics. DOI: 10.3389/fnbot.2021.730965.