Track objects in video with SORT and OC-SORT

In this video, Lee Clement, Machine Learning Engineer at Roboflow, demonstrates how to easily integrate multi-object tracking into your computer vision pipelines using Roboflow Workflows. Roboflow Trackers is a plug-and-play, open-source (Apache 2.0) library that provides clean, modular implementations of popular tracking algorithms like SORT, ByteTrack, and OC-SORT. Lee explains the differences between the tracking algorithms, outlining how simple trackers like SORT are optimized for high-throughput in stable environments, while OC-SORT is built to handle non-linear motion and occlusions. He then provides two live demonstrations showing how to use these trackers as drag-and-drop building blocks in Roboflow Workflows. First, you'll see how to build an industrial tracking pipeline from scratch using SORT to count objects moving on a production line. Next, Lee demonstrates how to use OC-SORT to track a soccer ball during a fast-paced game with heavy occlusions. = Chapters = 00:00 Introduction: Tracking Objects in Video 01:30 Comparing Algorithms: SORT vs. ByteTrack vs. OC-SORT 04:05 Building a Workflows Pipeline: Counting Butter Packages with SORT 07:54 Testing the Pipeline and Running Inference with Serverless Video 09:27 Handling Occlusion: Sports Tracking with OC-SORT 11:28 - Tuning and Testing the Minimum IOU Threshold Parameter = Additional Resources = Roboflow Trackers on GitHub: https://github.com/roboflow/supervision Getting Started with Roboflow: https://docs.roboflow.com/ Roboflow Workflows: https://roboflow.com/workflows