YOLOWEN:An Efficient Multi-object Detection Model for Cigar Appearance Defects

An innovative YOLOWEN model has been introduced to automatically identify defects in full-leaf, handmade cigars, including bruises, holes, breaks, and tail breaks, which are critical quality concerns during production. This model enhances the YOLOv5 architecture by incorporating a WEN (Weighted ENhancement) module, which enriches the model's receptive field with a hierarchical structure. Additionally, the traditional C3 module has been swapped out for a fusion module, enhancing the model's ability to integrate features more effectively. The inclusion of a CBAM (Convolutional Block Attention Module) at the detection head's forefront ensures that the model prioritizes significant image features while downplaying less relevant areas. Preliminary experimental outcomes indicate that the YOLOWEN model outperforms the standard YOLOv5 by 1.7 percentage points in [email protected], achieving an impressive 96.40%. This advancement signifies that the YOLOWEN model is highly effective in detecting key cigar defects, thereby offering a robust solution for the intelligent quality control in cigar manufacturing.