Overview
Multi-Gate Image Analysis (MGIA) is an image analysis tool which uses machine learning to recognize good and bad patterns, exclude good patterns from the void count, and combine defect analysis from different acoustic images into one straightforward result based on customized failure criteria.
MGIA Segmentation
The AI/ML model is able to distinguish three classes of pixels:
- Background
- Chip marker (green)
- True defect (blue)
The model is trained that the calibration voids, with consistent size, shape, and placement, are considered “background” pixels despite having the same pixel value as defects.