Rubo12345/Directed-Research-on-Vehicle-Re-Identification

In this project, a novel framework is used from the reference mentioned in the README, which successfully encodes both geometric local features and global representations to distinguish vehicle instances, optimized only by the supervision from official ID labels. Specifically, given the insight that objects in ReID share similar geometric characteristics, a self-supervised representation learning technique is used to facilitate geometric features discovery. To condense these features, an interpretable attention module, with the core of local maxima aggregation instead of fully automatic learning, in used, whose mechanism is completely understandable and whose response map is physically reasonable.

Python