Defect detection is a crucial task in screen-printed circuit (SPC) production, where image processing method based on deep learning is often used. This field frequently encounters challenges, such as minute surface defects, a large number of model parameters, and high computational complexity. To address these challenges, a self-made SPC defect data set and an enhanced CAAB-YOLOv8n detection algorithm were developed. A CAD module was integrated into the backbone network to improve the model’s ability to detect bar-shaped features. In addition, the ASF feature fusion and RMT modules were combined to construct the ASF-CR neck structure, which enhances the model’s capability to detect small, localized defects. To expedite inference speed, the DBB-Head reparameterization module was incorporated. Experimental results show that the enhanced algorithm achieves 88.4 $$\%$$ % accuracy, a mAP@50 of 90.2 $$\%$$ % , and a parameter count of just 33.27 million, with a detection speed of 35.2 frames per second. The real-time requirements for SPC defect detection are met by these findings. This work lays a solid theoretical foundation for subsequent defect traceability and the optimization of printing process parameters.

The dialogue summarization is necessary for information retrieval, and the training of abstract dialogue summarization models heavily rely on large amounts of labeled data. However, manual summarization of long dialogue is labor-costing and time-consuming. To solve this problem, this article proposes a data augmentation method for dialogue summary based on adaptive augmentation fusion (AAF), integrating the strengths of both Minor Perturbation Augmentation (MPA) and Semantic Reconstructive Augmentation (SRA) to balance model learning effectiveness and generalization capabilities. We first integrated existing enhancement methods to address the problem of insufficient annotated data for dialogue summarization. The experimental results on both the DialogSum and SAMSum datasets demonstrate that the AAF method achieves significant improvements in ROUGE scores under resource-constrained conditions, outperforming baseline approaches. Furthermore, it was validated that the selection of the amount of augmented data has a significant impact on model training results under resource-constrained conditions. We have publically released our code at https://github.com/alolke/AAF.