MCity Computer Vision Project

This dataset consists of still frames from phone videos taken at MCity, to aid with our object detection of things at MCity. We are only labeling four classes: pedestrians, vehicles, traffic lights, and traffic signs.

Pedestrians: any people in the images should have a bounding box from their head to their feet Vehicles: any cars in the images should have a bounding box spanning their entire body Traffic Signs: any traffic signs (i.e. stop signs, yield signs, street signs, etc., not banners or billboards) should have a bounding box that fits tightly around the sign area. Traffic Lights: any traffic lights (including backwards facing and/or sideways facing traffic lights, and traffic lights that are off) should have a bounding box that fits tightly around the whole frame of the traffic light (i.e. not just the bulbs or the inner area of the traffic light)

When in doubt, just use your best judgement! Also, the bounding boxes don't have to be absolutely pixel perfect, but try your best to have them line up reasonably well.