AI aims to transform a basic phone video into cycling performance metrics that serious cyclists invest thousands of dollars in.

      Cycling can be a costly hobby, especially for those who become more dedicated. As cyclists invest in measuring their performance, expenses can accumulate significantly with items like instrumented pedals, bike computers, heart-rate sensors, and other equipment that enhances the riding experience.

      Researchers at La Trobe University (via TechXplore) are investigating whether a common device that most people possess could provide a surprisingly advanced solution for tracking cycling metrics. This isn't a specialized tool; it's a small device that fits in your pocket—your smartphone. At the Holsworth Biomedical Research Centre, researchers are creating AI models that can estimate the forces exerted on the pedals by using motion captured on video. Cyclists can simply film themselves using their phones and obtain biomechanical data that would typically require specialized instruments.

      AI monitors each pedal stroke

      Aliya Barnwell / Digital Trends

      The system employs deep learning to link observed movements of the cyclist to the actual forces recorded at the pedals. For this study, the model was trained using synchronized lab recordings, which consist of video footage of the cyclist in motion and another set capturing the actual forces applied with each pedal stroke. The AI learns to predict these forces based solely on the movement.

      Initial results are already showing potential. An earlier study by La Trobe researcher Rodrigo Bini utilized recurrent neural networks to forecast three-dimensional pedal forces as well as forces and power at lower-limb joints. This research indicated correlations between the measured and predicted data ranging from 0.79 to 0.96, although the accuracy varied depending on the specific force or joint measurement. Thus, this new project might simplify scaling such analysis beyond the laboratory.

      Your phone won’t replace a power meter immediately

      Digital Trends

      This research is still in its infancy, and La Trobe is not launching an app anytime soon. Bini predicts that within the next five to ten years, smartphone applications could enable nearly anyone to record their movements and estimate cycling forces. The research team is currently enhancing and validating its dataset, with initial findings expected at the International Society of Biomechanics Conference next year.

      Previously, we have tested smart pedals, such as the Garmin Vector 3, which provide data on when and where force is applied during a pedal stroke. However, obtaining such detailed data has normally required specialized sensors and significant expenses. Utilizing a phone camera to approximate this technology could make valuable cycling biomechanics accessible to many more cyclists. With AI also being used in fitness coaching, we are likely to see technology increasingly fine-tuning various physical aspects of our lives.