Uploaded May 2026 | Updated November 2026, 2 days ago
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: github.com/roboflow/supervision
Getting Started with Roboflow: docs.roboflow.com
Roboflow Workflows: roboflow.com/workflows
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: github.com/roboflow/supervision
Getting Started with Roboflow: docs.roboflow.com
Roboflow Workflows: roboflow.com/workflows










