DeepSORT

DeepSORT strengthens motion-based tracking with visual identity embeddings, leading to better results in crowded scenes. Appearance information can reduce identity confusion when similar objects move through the scene. It is therefore widely used in person tracking, retail analytics, and security scenarios. It is one of the most effective practical examples of the tracking-by-detection paradigm.