Learning Spatial and Temporal Attention for AI-driven Virtual Sports Trainers

Fitness apps are widespread, used by professionals and casual users alike. Existing apps only give examples, they don’t correct. The proposed research closes this feedback loop by using the video camera on mobile phones as an analysis tool to judge and correct the execution of common gym exercises. This is particularly important for weight lifting, where inappropriate poses and motions commonly lead to injury. Furthermore, small improvements in form can lead to large improvements in efficiency. Besides huge practical impact potential, the developed approaches will extend the video classification literature to further increase accuracy and applicability to everyday scenarios.

Faculty Supervisor:

Helge Rhodin

Student:

Partner:

Flex AI

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

The University of British Columbia

Program: