Turbine Defect Detector - AI/ML

Sponsored by Collins Aerospace, me and a team of 3 students were tasked to build a composite structures defect detector aka airplane turbines. We did this by creating a CNN in a cloud-based environment that would take images of damaged and undamaged images of turbines and highlight the damaged ones.

Key Points:
• Achieved validation accuracy above 90% & 95% efficiency rating for ML model using dataset of 18.1k pictures
• Using Google Collab, delivered cloud-based solution: Turbine Defect Detector using Machine Learning Algorithms; Collaborated with stakeholders to create 1-year roadmap; product still being actively sourced