A new La Trobe University research project aims to bring elite-level cycling performance analysis to amateur athletes and coaches using nothing more than a smartphone camera.
Researchers from the Holsworth Biomedical Research Centre are developing AI models that use video data to estimate the forces cyclists apply to their pedals.
The models are still in the early stages of development, but researchers say they have the potential to improve performance, decrease injury risk and replace expensive equipment that is currently confined to specialist biomechanics laboratories.
More than 1.4 million Australians ride a bike each day, according to data from the 2025 National Cycling Participation Survey.
Previous research led by La Trobe's Associate Professor Rodrigo Rico Bini found knee pain is one of the most common overuse issues affecting cyclists.
Dr Bini said the research would use deep learning, a type of AI that can learn complex relationships, to assess a cyclist's movements and the force applied to the pedals.
"Motion data tells us how the body is coping with the demands placed on it by the bike," Dr Bini said.
"Awkward positions can indicate increased stress on muscles, tendons and bones, which may contribute to injury.
"Motion can now be captured easily using smartphones, so we're exploring whether AI can use that information to accurately estimate pedal forces."
The model is trained using synchronised laboratory measurements of pedal forces and video-based motion data, allowing it to learn and predict force throughout each pedal stroke.
While the current study uses video data, the approach could also work with other sources of motion information, including wearable sensors or marker-based motion capture systems.
The project builds on an earlier collaboration with researchers in Spain, which demonstrated that AI models could accurately estimate cycling forces from motion data. This second stage will test tools which could make the technology easier to scale and use in real-world settings.
Dr Bini is also collaborating with long-term research partner Associate Professor Felipe Arruda Moura, from State University of Londrina in southern Brazil, to strengthen the dataset used to train and validate the AI models.
Dr Bini said the technology could eventually help cyclists improve performance and identify injury risks without the need for specialist testing.
"In the next five to 10 years, smartphone apps could allow almost anyone to capture video and estimate cycling forces," Dr Bini said.
Dr Bini will present preliminary findings from the project at the International Society of Biomechanics Conference in Sydney next year, where researchers from around the world will showcase the latest advances in human movement science.
Explore how we're shaping the future of AI across research, education and partnerships La Trobe AI Institute | La Trobe University.
