StairNet

Computer vision can be used in robotic exoskeleton control to improve transitions between locomotion modes through the prediction of future environmental states. We developed StairNet to focus specifically on stair recognition due to the potential safety implications and the theoretical risk of injury resulting from environmental misclassification during stair ascent. The dataset was developed using the “ExoNet” database – the largest and most diverse open-source dataset of wearable camera images of walking environments. StairNet contains 515,452 labelled images from six of the twelve original ExoNet classes. These images were carefully reclassified into four classes which use novel definitions created from a computer vision perspective with the goal of increasing the accuracy of the cutoff points between classes within the dataset. Additionally, the dataset was manually parsed numerous times during annotation to reduce misclassification errors and remove images with large abstractions from the dataset.