Manually inspecting micrographs for defects and classifying them is very time consuming and typically costs over £100k per annum.
80% in human operator involvement time saved and thousands of classifications are obtained within 2 hours.
- Collate micrographs and data associated with them
- Develop Convolutional Neural Network model
- Develop a system to classify micrograph images as containing defects or not.
- Validate results. Trials showed over 90% accuracy in classifying micrographs