| Abstract |
Logo detection is a crucial task for industrial cyber-physical systems in smart cities, which relies on accurate and efficient analysis of urban environments. In this article, we proposed SC-YOLO, a robust logo detection framework that utilizes an optimized YOLO architecture with several key enhancements. Specifically, this framework encompasses a focus module that enables lossless downsampling and expands the input image channel to enrich feature representation. In addition, we incorporate an attention mechanism that enhances the model's ability to extract and locate critical regions in an image. To facilitate iteration and improve accuracy, we introduce the SIoU loss function that considers angle, distance, and shape costs between targets. The experiments on the LogoDet-3 K dataset demonstrate that SC-YOLO achieves an mAP of 70.84%, outperforming other state-of-the-art deep detection models, including YOLOv7. Overall, our proposed method has a wide range of potential applications in autonomous driving, intelligent transportation, smart city advertising research, and smart city commercial branding, enabling more efficient and safer smart cities. |
