Machine learning/computer vision

Section 3

Machine learning/computer vision

Best practice knowledge for the principles associated with the development of machine learning (video or biomechanical data) models and sports-specific applications of computer vision. Understanding the theoretical concepts and appropriate application of different types of machine learning models (supervision levels) and considerations for each components of machine learning (Representation, Evaluation and Optimisation).  Appropriate and competent knowledge of computer vision theory needed for integration into development of bespoke video machine learning models.

• Base level – practitioner is familiar with basic theory associated with the appropriate application and limitations associated with machine learning including data integration, selection, pre-processing; model selection; training/testing phases of model development; and data interpretation. Able to conduct executable machine learning models (eg. open-source pose estimation models)
• Intermediary level – practitioner is capable of incorporating machine learning scripts and appropriate algorithms in relevant programming language (eg. Python) to develop machine learning routines. Competent in basics operation of appropriate commercial or open source machine learning programs (such as DeepLabCut).
• Advanced level – practitioner has recognised high-level expertise in utilising and developing machine learning models using best practice principles to allow for bespoke applications (such as development of sports/skill-specific video machine learning pose estimation models for sports biomechanics testing applications).