Person re-identification using part-based convolutional baseline

Abstract

Person re-identification is the process of identifying of a person previously identified. This has become an area of increasingly popular research due to its application in the public security. In comparison to other machine learning that also involve searching for object, person re-identification is of higher difficulty due to the various variation that can happened in real-world condition. The variation consists of brightness, image resolutions, the point of view and the obstruction of body parts during capture. With these variations in place, the project’s objective is to create a person re-identification system that can correctly predict the input image (query) from within a pool of data image. This project will focus on the Part-based Convolutional Baseline and refined part pooling.