SDES-YOLO: A High-Precision and Lightweight Model for Fall Detection in Complex Environments

This paper presents SDES-YOLO, a high-precision and lightweight model for fall detection in complex environments, building upon the YOLOv8 framework. By integrating a multi-scale feature extraction pyramid (SDFP), occlusion-aware attention mechanism (SEAM), an edge and spatial information fusion module (ES3), and a WIoU-Shape loss function, SDES-YOLO significantly enhances detection accuracy in challenging conditions such as varying lighting, occlusions, and complex human postures. With only 2.9M parameters and 7.2 GFLOPs of computation, SDES-YOLO achieves a mAP@0.5 of 85.1%, outperforming YOLOv8n by 3.41%, while reducing parameters and computation. This model combines efficiency and precision, making it highly effective for fall detection even in resource-constrained environments.

Xuewen Wang, Qingzhan Zhao, Ping Jiang, Yuchen Zheng, Limengzi Yuan, Panli Yuan. LDS-YOLO: A lightweight small object detection method for dead trees from shelter forest. Comput. Electron. Agric. 198: 107035 (2022)