Research on Lightweight Object Detection Based on Improved YOLOv8n for Experimental Scenarios

Lightweight object detection algorithms, represented by YOLOv8n, have achieved significant optimization in terms of model size and parameter count. However, the neck network of YOLOv8n is not efficient at fusing contextual information, and the backbone network's feature extraction capabilities are limited. This results in challenges such as missed and incorrect detections in complex images, occluded objects, or scenes with significant background noise, making it difficult to meet the requirements of end-to-end industrial deployments. To address these issues, this paper proposes a lightweight object detection algorithm based on an enhanced YOLOv8n architecture. By integrating FasterBlock, GhostConv, and BiFPN modules, the model's detection performance and feature extraction capabilities are significantly improved, while the overall model size is reduced. Additionally, we constructed a dataset based on experimental scenarios to validate the effectiveness of the proposed model. Experimental results show that the modified model improves detection accuracy by approximately 7% compared to the original YOLOv8n, reduces the parameter count from 3.01 million to 1.73 million. This significantly enhances detection accuracy while substantially reducing the model's size. The improved YOLOv8n model not only demonstrates superior detection performance but also effectively reduces model complexity and resource consumption. This research lays a solid foundation for future applications of AI-driven automated scoring systems in experimental environments and provides a new direction for end-to-end industrial deployment.