Skip to main content

Song Yan: AI computer vision helps to estimate body shape and track moving objects

Tampere University
Date4.11.2022 12.00–16.00 (UTC+2)
LanguageEnglish
Entrance feeFree of charge
Man looking at camera with a cap hat on. Trees and water on the background.
In his doctoral dissertation, Song Yan collected new datasets from real human scans by developing methods that provide state-of-the-art results. Yan’s study is conducted on two research lines: the first line focuses on using AI computer vision tool for measuring the human body. The other line emphases computer vision for tracking moving targets in videos.

Body shape analysis, including pose and volumetric 3D shape, has many applications in healthcare, online shopping, ergonomic design, human-computer interaction, and sports. With the help of modern artificial intelligence (AI) methods and tools, it is possible to provide accurate estimates of human pose and body shape.

AI computer vision tool is easy to use”, says Yan. “Simply taking a photo of yourself with a mobile phone camera is enough for the AI tool to estimate your body shape. This function helps you to find well-fitting clothes from online stores.”

AI tool can also be used for investigating the functionality of diets and exercise programs and detecting certain illnesses. Accurate estimation of a body shape is based on artificial intelligence that learns body shapes from training samples taken from volunteered subjects.

Yan worked in close collaboration with several Finnish companies that have further developed the methods into commercial products. The AI computer vision tool can be tested on Sizey Ltd’s website.

Use of the AI computer vision tool in object tracking

Object tracking aims to follow a moving object, such as a car, human, or animal in difficult conditions. Yan developed new datasets and state-of-the-art methods, particularly following objects in depth videos by using both depth and conventional video cameras. Using these methods, for example drones can follow objects, and provide them security.

As a part of his Ph.D. thesis, Yan participated in the most important benchmark of object tracking, Visual Object Tracking Challenge Workshop (VOT) in 2022. In VOT all the leading groups from the field compete against one another. The results were announced in the annual workshop of the challenge in Tel Aviv, Israel. You can see the results and winners on VOT’s website

“I believe that these technologies will be found from many applications in the next few years. I am happy that I was able to contribute science and learn the modern AI techniques during my Ph.D. thesis”, Yan remarks.

Yan now works for the Honor mobile phone company where he develops new AI imaging technologies for future mobile phones.

The doctoral dissertation of MSc (Tech) Song Yan in the field of Computer Vision titled Vision and Depth Based Computerized Anthropometry and Object Tracking will be publicly examined in the Faculty of Information Technology and Communication Sciences at Tampere University at 12 o'clock on Friday, 4 November 2022. The Opponents will be Professor Miguel Pordallo Lopez from the University of Oulu, Finland, and Doctor Roman Pflugfelder from the Technical University of Munich, Germany. The Custos will be Professor Joni Kämäräinen from the Faculty of Information Technology and Communication Sciences at Tampere University.

The doctoral dissertation is available online.

Photo: Wenyan Yang