Semiconductor wafer chip defect detection algorithm based on improved YOLOV8

In this paper, target detection of semiconductor wafer chip defects is realized based on the improved YOLOV8 network. The ResNet50 backbone network is used to replace the CSPDarknet backbone network in the original algorithm, which effectively solves the gradient vanishing problem of the deep network and provides more stable feature extraction capability. By using the improved CSPPC module instead of the original C2f module, the computational redundancy is reduced while preserving the critical information circulation path, improving the feature representation capability and controlling the computational cost. The GD mechanism is fused in the neck network to enhance the multi-scale target checking detection capability of the model. Finally, the accuracy of the improved YOLOv8 algorithm is demonstrated experimentally.