MVAD-Net: Learning View-Aware and Domain-Invariant Representation for Baggage Re-identification

Baggage re-identification (ReID) is a particular and crucial object ReID task. It aims to only use the baggage image data captured by the camera to complete the cross-camera recognition of baggage, which is of great value to security inspection. Two significant challenges in the baggage ReID task are broad view differences and distinct cross-domain characteristics between probe and gallery images. To overcome these two challenges, we proposeMVAD-Net, which aims to learn view-aware and domain-invariant representation for baggage ReID by multi view attention and domain-invariant learning. The experiment shows that our network has achieved good results and reached an advanced level while consuming minimal extra cost.