Person re-identification: Implicitly defining the receptive fields of deep learning classification frameworks

Ehsan Yaghoubi, Diana Borza, S. V. Aruna Kumar, Hugo Proença

Abstract: Highlights•A data-oriented technique for person Re-ID, helping CNNs to find the accurate receptive fields.•Synthesizing n2−n pedestrian samples from n available samples.•A holistic, parameter-free, and on-the-fly person data augmentation technique.