Optimizing the visual effects of advertisement design using image analysis technology based on the YOLOv4 target detection algorithm

Abstract

The lack of efficient quantitative analysis methods for visual elements in advertisement design often results in suboptimal visual impact and ineffective communication of key information. To address this challenge, this study employs image analysis technology based on the YOLOv4 (You Only Look Once v4) object detection algorithm to enhance the visual effectiveness of advertisement design. First, a diverse set of advertisement samples is collected, and image augmentation techniques are applied to expand the dataset. The YOLOv4 model is then trained using the CSPDarknet53 (Cross-Stage Partial Darknet53) network for feature extraction, integrating multi-scale prediction to improve detection accuracy. The trained model is subsequently utilized to identify and analyze visual elements within advertisements, assessing their spatial distribution and layout characteristics. Based on these insights, an automated optimization framework is developed to refine design elements such as color, size, and layout proportions, with the results visualized for comparative analysis. Experimental results demonstrate that the proposed method achieves a detection precision of 98.1% and a recall rate of 96.8% for product elements. The optimized print advertisements exhibit a focus concentration of 0.84 and an element distribution density of 0.87, significantly enhancing design clarity and effectiveness. This approach not only improves the efficiency and precision of quantitative analysis in advertisement design but also offers a practical and scalable solution for optimizing visual communication strategies.