Adaptive Graph Attention Network in Person Re-Identification

Abstract

This paper approaches a significant problem in computer vision: re-identifying a person when having groups of people. Re-identifying (Re-ID) by group context is a new direction for improving traditional single-object Re-ID task by additional information from group layout and group member variations. Adding new improvements in the graph convolution layer structure or using more powerful theories boosts the model’s accuracy. To handle these issues, we proposed to leverage the information of group objects: person and subgroups of two or three people inside a group image. We have reorganized the load data structure to improve training efficiency. Organizing the data is based on the relational representation of the central node, and the observed nodes further incorporate their features extracted through the backbone is Resnet. We propose a graph convolution model that uses the Selu activation function to study this data. The key challenge in implementing is to define the optimal group-wise matching using adaptive graph attention based on a graph convolution network modified and training techniques. However, we only did this training for a limited amount of time with 20 epochs per training due to resource limitation. The key challenge in implementing is to define the optimal group-wise matching using adaptive graph attention based on a graph convolution network modified and training techniques. However, we only did this training for a limited amount of time with 20 epochs per training due to resource limitation.