Methods, systems, and apparatus, including computer programs encoded on computer-readable media, for obtaining a first image from a first sensor; detecting a first object within the first image; extracting discriminative features of the first object into a first feature vector; obtaining a second image from a second sensor; detecting a second object within the second image; extracting discriminative features of the second object into a second feature vector; and determining, based on a comparison between the first feature vector and the second feature vector, that the second object is same as the first object.