This paper presents a simple and efficient baseline for person re-identification (ReID) using deep neural networks. The authors collect and evaluate various effective training tricks from existing literature, combining them to achieve high performance with only global features. The modified baseline, incorporating six key tricks, achieves 94.5% rank-1 accuracy and 85.9% mAP on the Market1501 dataset. The paper also explores the impact of image size and batch size on model performance and provides a strong baseline for researchers and industry practitioners. The code and models are available on GitHub. The contributions include a comprehensive evaluation of training tricks, a strong ReID baseline, and insights into the effectiveness of these tricks. The authors aim to promote the development of ReID research by providing a robust starting point and encouraging further exploration of effective techniques.This paper presents a simple and efficient baseline for person re-identification (ReID) using deep neural networks. The authors collect and evaluate various effective training tricks from existing literature, combining them to achieve high performance with only global features. The modified baseline, incorporating six key tricks, achieves 94.5% rank-1 accuracy and 85.9% mAP on the Market1501 dataset. The paper also explores the impact of image size and batch size on model performance and provides a strong baseline for researchers and industry practitioners. The code and models are available on GitHub. The contributions include a comprehensive evaluation of training tricks, a strong ReID baseline, and insights into the effectiveness of these tricks. The authors aim to promote the development of ReID research by providing a robust starting point and encouraging further exploration of effective techniques.