Developing an AI model to identify rust for ServiceEcho

Advancing artificial intelligence and augmented reality technologies to improve the detection and assessment of rust on metal surfaces

Dushyant Puri, Mark Buchner, and Mahmud Ashrafizaadeh, SMART Centre

About the project

This project builds on earlier research focused on using artificial intelligence to identify rust in real time. The work introduced new tools to simplify image collection, annotation and model improvement, creating a more efficient process for developing and refining machine learning systems.

The challenge

Machine learning models require large volumes of accurately labelled data to achieve reliable performance. Collecting and annotating images across different environments can be time-consuming and resource-intensive, creating a barrier to developing effective computer vision solutions.

Approach

• Enhanced an existing rust-detection system with new image collection and annotation capabilities
• Developed tools to support image capture, review and editing workflows
• Created a process for refining machine learning models using user-generated feedback
• Integrated mobile and web-based technologies to streamline data management
• Explored opportunities to combine artificial intelligence with augmented reality experiences

Results

• Developed the AnnoVision system to support image collection and annotation
• Improved the workflow for generating and refining training data
• Enabled users to review and adjust model-generated annotations
• Established a framework for continuously improving model accuracy
• Created a foundation for future augmented reality and artificial intelligence applications

Why it matters

• Supports the development of more accurate computer vision systems
• Reduces the effort required to collect and label training data
• Demonstrates practical applications of artificial intelligence and augmented reality technologies
• Creates opportunities for enhanced inspection and maintenance processes
• Establishes a scalable approach for future machine learning development

Get started with your next project

Funding for this project was provided by the Southern Ontario Network for Advanced Manufacturing Innovation (SONAMI).