Published Paper Details:

This study presents a groundbreaking approach to enhance the accuracy of the YOLOv8 model in object detection, focusing mainly on addressing the limitations of detecting objects in varied image types, particularly for small objects. The proposed strategy of this work incorporates the Context Attention Block (CAB) to effectively locate and identify small objects in images. Furthermore, the proposed work improves the feature extraction capability without increasing model complexity by increasing the thickness of the Coarse-to-Fine(C2F) block. In addition, Spatial Attention (SA) has been modified to accelerate detection performance. The enhanced YOLOv8 model (Namely YOLOv8-CAB) strongly emphasizes the performance of detecting smaller objects by leveraging the CAB block to exploit multiscale feature maps and iterative feedback, thereby optimizing object detection mechanisms. As a result, the innovative design facilitates superior feature extraction, "especially the weak features," contextual information preservation, and efficient feature fusion. Rigorous testing on the Common Objects in Context (COCO) dataset was performed to demonstrate the efficacy of the proposed technique. It is resulting in a remarkable improvement over standard YOLO models. The YOLOv8-CAB model achieved a mean average precision of 97% of detecting rate, indicating a 1% increase compared to conventional models. This study highlights the capabilities of our improved YOLOv8 method in detecting objects, representing a breakthrough that sets the stage for advancements in real-time object detection techniques.