CDD-YOLOv8: A Small Defect Detection and Classification Algorithm for Cigarette Packages

In order to ensure the quality of cigarette products, it is significant to conduct real-time detection and classification for defects on cigarette packages in high-speed assembly line. Most industrial anomaly detection methods can only detect anomalies by modeling normal data distribution, lacking the ability to realize the fine-grained classification of anomalies, which is indispensable for reducing defective products and thus lower production costs. To solve the problem of detecting and classifying small defects in cigarette packaging, we introduced a Cigarette Defect Detection YOLOv8 algorithm (CDD-YOLOv8). By adding a small object detection head and Convolutional Block Attention Module(CBAM), our proposed model achieved better utilization of multi-level features to assist in small defect detection, contributing to better detection performance in our cigarette dataset.