- Fri 20 November 2026
- news
- Jason K. Moore
- #bicycle, #engineering, #camera, #machine learning, #marker-less motion capture
Haye successfully defended his MSc Thesis title Development of a 3D bicycle keypoint detection method for extracting kinematics data from monocular video data on August 24, 2026.
Haye created synthetic monocular camera footage of bicyclist avatars navigating a scene. He then developed a pipeline which first extracted the key points of the bicycle's geometry from the 2D footage. The key points were then "lifted" to to three dimensional space using a second machine learning technique that leveraged the dynamic nature of the rigid body system's differential equations. Finally, he extracted commonly measured variables like steer and roll angle and showed that he could accurately extract these variables from the monocular footage alone.
Example showing the extraction of the 3D bicycle geometry from video frames.
Haye was supervised by Benjamín González, Ajay Seth, and Jason K. Moore. Everyone at the bicycle lab is very proud of Haye and wishes him the best with his upcoming PhD in Groningen.