Person Re-Identification Using Integrated Attention and Graph Neural Network

Person re-identification (P-ReID) remains challenging due to occlusions, background clutter, and variations in camera viewpoints. While deep learning techniques have significantly improved performance, feature alignment and robustness persist. This paper proposes a hybrid deep learning framework integrating Convolutional Neural Networks (CNNs), Position Attention Module (PAM), Efficient Channel Attention (ECA), and Graph Neural Networks (GNN). The combination enhances spatial awareness, channel-wise feature representation, and relational modeling between images. Experiments conducted on the Market-1501 dataset demonstrate an improved mean Average Precision (mAP) of 82.15% and Rank-1 accuracy of 90.5%. This hybrid approach advances P-ReID methodologies, offering a foundation for future deep-learning research.