Improved YOLOv8 Model Based on Multi-Scale Context Aggregation Architecture

With the rapid development of deep learning, numerous deep learning-based object detection algorithms have emerged. Upon studying the YOLOv8 model, it was found that there are certain issues with its detection of small objects, and its capability to handle multi-scale targets and complex scenes needs improvement. Additionally, its feature extraction ability requires enhancement. To address these issues, this paper proposes an improved YOLOv8 model based on expanded convolutional modules and additional detection heads. Building upon the traditional structure of YOLOv8, new detection heads were introduced, and expanded convolutional modules were integrated into the model to widen the receptive field of convolutional layers, thereby extracting richer image features and enabling the network to better understand contextual information about targets. Performance testing was conducted on the PASCAL VOC2007 public dataset, and comparative experiments with classical object detection algorithms were performed. The experimental results demonstrate that the enhanced YOLOv8 model with expanded convolutional modules and additional detection heads achieves higher detection accuracy.