Large-scale benchmark for vehicle Re-Id in the real-world urban surveillance
To facilitate the research of vehicle re-identification (Re-Id), we build a large-scale benchmark dateset for vehicle Re-Id in the real-world urban surveillance scenario, named “VeRi”. The featured properties of VeRi include: It contains over 50,000 images of 776 vehicles captured by 20 cameras covering an 1.0 km^2 area in 24 hours, which makes the dataset scalable enough for vehicle Re-Id and other related research. The images are captured in a real-world unconstrained surveillance scene and labeled with varied attributes, e.g. BBoxes, types, colors, and brands. So complicated models can be learnt and evaluated for vehicle Re-Id. Each vehicle is captured by 2 ∼ 18 cameras in different viewpoints, illuminations, resolutions, and occlusions, which provides high recurrence rate for vehicle Re-Id in practical surveillance environment. It is also labeled with sufficient license plates and spatiotemporal information, such as the BBoxes of plates, plate strings, the timestamps of vehicles, and the distances between neighbouring cameras.