Machine Learning and Computer Vision for Accurate Foot Pose Estimation for Clinical and Sporting Applications

Overview

Develop machine learning methods for accurate foot pose estimation from multiple RGB cameras using markerless motion capture techniques. Despite recent progress in human pose estimation, accurately capturing fine-grained foot motion remains challenging due to occlusion, complex articulation, and limited annotated datasets. Improving foot pose estimation is essential for high-fidelity biomechanical analysis and downstream applications in healthcare and performance science. This project addresses these challenges by designing robust, data-driven models that integrate domain knowledge from biomechanics with advances in computer vision and machine learning.

Research engagement

You will engage with applied research in computer vision and machine learning for studying human movement, with particular emphasis on the foot and ground contact. This will include: a review of review and evaluation of existing pose estimation techniques, data preprocessing, feature extraction, model implementation, and performance evaluation.

Research activities

You will work with the supervisors and, where appropriate, academic staff who have experience with biomechanics and machine learning Activities may include:

• Apply image inpainting to remove motion capture markers
• Evaluate existing pose estimation methods
• Develop foot-specific machine learning models
• Registration of human foot models with ground-truth high resolution 3D scans
• Model foot and ground contact from RGB images

Research skills

You will gain experience in engineering research and development, and how to translate this research to address real-world problems, particularly for clinical applications. You will develop skills needed for research, such has how to thoroughly evaluate existing research, identify important gaps, and how to develop solutions to these problems.

Outcomes

Develop improved machine learning models for accurate foot pose estimation from multiple calibrated and simultaneously captured RGB cameras.

Skills and experience

• Machine Learning concepts
• Software Version Control and Software Engineering best practices
• Computer vision/image processing concepts

Start date

2 November, 2026

End date

19 February, 2027

Location

QUT Gardens Point Campus

Keywords

• Machine Learning
• Pose Estimation
• Biomechanics

Contact

Akila

a.thondilege@qut.edu.au