CAE-YOLOV8: Occlusion Object Detection Based on Improved YOLOv8

Occlusion recognition plays a vital role in smart manufacturing and industrial automation. This paper proposes an enhanced YOLOv8-based model designed to accurately recognize and locate occluded objects in industrial environments, assisting robots in precise grasping tasks. The model integrates the AKConv module and the Effciou loss function, which improve feature extraction and bounding box regression, especially in complex and highly occluded scenarios. Experimental results confirm that the modified YOLOv8s_Effciou model achieves a balanced performance across different levels of occlusion, showing notable robustness in challenging environments with high occlusion.