Object Detection for E-scooters

ScooterDet dataset comprises 2013 images encompassing 11 traffic object classes, i.e., “person”, “car”, “truck”, “bus”, “traffic light”, “fire hydrant”, “stop sign”, “bench”, and “scooter”, with a total of 11,011 bounding box annotations. The dataset is collected using Tobii Pro Glasses 2 mounted on a Segway NineBot scooter, traversing from the University of Virginia campus to the urban area of Charlottesville, VA, USA. Image frames are extracted from the recorded videos captured by the Tobii Pro Glasses, and subsequent data cleaning procedures are employed to eliminate low-quality images and those lacking relevant objects of interest.