Deep SORT is an implementation of Simple Online and Realtime Tracking with a Deep Association Metric, extending SORT by incorporating a deep appearance descriptor for data association in multi-object tracking. The repository provides installation steps, example usage, and scripts to run the tracker, generate deep appearance features for person re-identification, and evaluate results on MOT challenge datasets. It relies on a CNN-based appearance descriptor to compare bounding box appearances across frames, enabling robust, online tracking of pedestrians and other objects.
AI named DeepSORT from March 2026 to July 2026.
AI named DeepSORT alongside Amazon Rekognition, Azure AI Video Indexer, and Google Cloud Video Intelligence
None of these answers cited github.com. AI leaned on aws.amazon.com and learn.microsoft.com.
github.com/nwojke/deep_sort 0%Other sites 100%