adesso Blog

Object recognition is widespread and is used in various applications such as recognising people, cars and so on. However, it lacks the intelligence to determine whether different images in video sequences are the same recognised object. Therefore, object tracking is crucial for many machine vision applications. From analysing flows of people and traffic to analysing movement patterns in sporting activities, object tracking plays an indispensable role in gaining valuable insights.

With advances in artificial intelligence and machine learning, object tracking algorithms such as Multiple Object Tracking (MOT) and Single Object Tracking (SOT) are becoming increasingly intelligent, enabling more accurate and faster object tracking. While MOT tracks multiple objects simultaneously and is therefore indispensable in busy scenarios such as traffic or team sports, SOT focuses on a single object. Given this variety of use cases, the question arises as to which object tracking algorithm is best suited for a particular use case. To answer this question, we conducted an experiment, the results of which I would like to present in this blog post.

The experiment

In this experiment, different methods for tracking multiple objects were evaluated using various metrics, with the MOT17 competition serving as a benchmark.

Metrics at a glance


The experiment showed that the algorithms delivered different results in different use cases, making it clear that there is no "one size fits all" solution. Some trackers such as SORT showed remarkable speed, while their accuracy was significantly affected in scenarios with frequent occlusions and fast object movements. Other methods where a deep learning model is integrated into the tracker, such as DeepSORT, can handle occlusions very well but have problems with fast object movements. Other methods such as ByteTrack are promising in dealing with overlapping trajectories, but show limitations in maintaining consistent track identities over longer periods of time.

Result of the various tracker algorithms

MOT17 raw (source:

ByteTrack Output


Object tracking in video is an important field with many potential applications. The results of this experiment provide a valuable insight into the performance of different object tracking methods and suggest that the choice of the right method depends heavily on the specific requirements of the use case.

Would you like to find out more about exciting topics from the world of adesso? Then take a look at our previous blog posts.

Would you like to find out more about AI and how we can support you? Then take a look at our website. Podcasts, blog posts, events, studies and much more - we offer you a compact overview of all topics.

Find out more on our website

Picture Aynur Amirfallah

Author Aynur Amirfallah

Aynur Amirfallah has been working at adesso for several years as a data scientist specialising in deep learning and computer vision. In her current role in the Competence Centre Data Science, she is responsible for advising customers on AI solutions and their implementation.

Our blog posts at a glance

Our tech blog invites you to dive deep into the exciting dimensions of technology. Here we offer you insights not only into our vision and expertise, but also into the latest trends, developments and ideas shaping the tech world.

Our blog is your platform for inspiring stories, informative articles and practical insights. Whether you are a tech lover, an entrepreneur looking for innovative solutions or just curious - we have something for everyone.

To the blog posts

Save this page. Remove this page.