YOLO-WS: A Novel Method for Webpage Segmentation

To address the limitations of traditional heuristic and machine learning-based webpage segmentation algorithms in feature extraction performance and efficiency, we propose a webpage segmentation method based on deep learning object detection. Specifically, we propose a webpage segmentation method named YOLO-WS based on the YOLOv5 model. We optimized and improved the YOLOv5 model’s network structure, loss function, and post-processing for webpage segmentation tasks, and then use transfer learning to train YOLO-WS on the improved model. Experimental results show that YOLO-WS achieves good performance in web page segmentation tasks.