Mixed Wafer Defect Model

Overview

Implementation of a PyTorch-based model for identifying mixed defect types on semiconductor wafers. Wafer defect maps contain complex, overlapping defect signatures that are difficult to classify with standard single-label approaches. Addresses the mixed-label problem where multiple defect patterns coexist on a single wafer. Built and trained in a Jupyter notebook environment with model weights saved for reuse. Targets the quality assurance problem in semiconductor manufacturing where accurate defect classification directly impacts yield analysis and process control.