SCW-YOLO: An Improved Algorithm for Fall Detection Based on Deep Learning

Fall detection plays a crucial role in fields such as smart surveillance and healthcare applications. However, existing fall detection techniques suffer from low accuracy and poor adaptability to specific scenarios. To address these issues, we propose a fall detection system based on an improved YOLOv8n algorithm. First, we integrate the C2f module with SCConv, restructured into the C2f_SCConv module, to reduce spatial and channel redundancy in convolutional neural networks, thereby improving efficiency and accuracy. Second, we utilize the CARAFE upsampling operator to replace the original upsampling operator in YOLOv8, enhancing the model's ability to recognize details and local features, thus improving robustness and adaptability across different scenes. Lastly, we adopt WIoU as the bounding box regression loss function, replacing the original CIoU, to enhance the model's ability to detect small targets and pose details. Experimental results demonstrate that compared to the YOLOv8n model, our method achieves a 2.3% improvement in mAP@0.5 for fall detection, indicating its effectiveness in fall detection tasks.