Partial person re-identification with convolutional neural network and attention model

Abstract

We address a partial person re-identification problem, where only a part of person is observed and full body images are presented to be matched. This partial person re-identification is more challenging problem than conventional person re-identification problem which only considers full body images of person. In order to solve this problem, we proposed end-to-end deep model which make use of convolutional neural network (CNN), ROI Pooling layer, and attention model. For evaluation for proposed model, we process CUHK03 data to make simulated data, p-CUHK03, and quantitatively evaluated proposed model.