Chip Surface Defect Recognition based on YOLOv11

Chip surface defect inspection is a critical process to manufacture semiconductors with excellent performance and stable integrated circuits. In this paper, we present a cutting-edge YOLOv11-based defect inspection model, which can support realtime processing and high-precision detection. The model can detect and classify three typical chip surface defects, i.e., cracks, ink marks, and broken areas, with high precision. Innovative additions like the C3K2 block, SPFF module, and C2PSA attention mechanism improve defect localization and classification to a large extent. The model is 0.99 accurate when IoU threshold is 0.5, and it surpasses state-of-the-art algorithms. This paper verifies the feasibility of deep learning-based defect inspection in industrial control and quality control.