Solving long-standing challenge in semiconductor manufacturing—a refined algorithm for detecting silicon wafer defects

https://techxplore.com/news/2024-09-semiconductor-refined-algorithm-wafer-defects.html

"automated detection, deep learning, SPD-Conv specialized convolutional neural network identifies minute irregularities... balances accuracy against detection speed...  Convolutional Block Attention Module added to model, sharpening focus on smaller defects that are often missed in manual inspection or by other algorithms... achieved a mean average precision of 92.5% and had a recall rate of 94.1%. It did this quickly, at a rate of 136 images per second, which is faster than earlier systems"