ExoNet Database: Wearable Camera Images of Human Locomotion Environments

Recent advances in computer vision and artificial intelligence have allowed researchers to develop environment recognition systems for lower-limb exoskeletons and prostheses. However, insufficient and private training datasets have impeded the widespread development and dissemination of image classification algorithms for environment recognition. To address these shortcomings, we developed ExoNet - the first open-source large-scale hierarchical dataset of high-resolution wearable camera images of human locomotion environments. Unparalleled in both scale and diversity, ExoNet comprises over 5.6 million images of different indoor and outdoor real-world walking environments, which were collected using a lightweight wearable smartphone camera system. Approximately 940,000 images in ExoNet were human-annotated using a 12-class hierarchical classification architecture. Available publicly through IEEE DataPort, ExoNet offers an unprecedented communal platform for training, developing, and comparing image classification algorithms for next-generation environment recognition systems. Beyond the control of lower-limb exoskeletons and prostheses, applications of ExoNet extend to humanoid and autonomous legged robotics.近年来，计算机视觉与人工智能领域的长足进展，使得研究者能够开发面向下肢外骨骼（lower-limb exoskeletons）与假肢（prostheses）的环境识别系统。然而，训练数据集的稀缺与私有化问题，却阻碍了环境识别用图像分类算法的大规模开发与推广部署。为解决上述痛点，我们构建了ExoNet——首个开源大规模层级化数据集，收录了人类移动环境下的高分辨率可穿戴相机图像。该数据集在规模与多样性上均具领先优势，包含超过560万张覆盖真实室内外步行场景的图像，采集自轻量化可穿戴智能手机相机系统。其中约94万张图像通过12类层级分类架构完成了人工标注。ExoNet通过IEEE数据港（IEEE DataPort）面向公众开放，为下一代环境识别系统的图像分类算法训练、开发与对比提供了前所未有的公共研究平台。除下肢外骨骼与假肢的控制场景外，ExoNet的应用场景还可拓展至人形机器人与自主腿式机器人领域。