LWP-YOLO: Detection of High-Precision Networks in Tennis Scenarios

In this paper, an improved YOLOv10 framework (LWP-YOLO) is proposed for enhancing the detection accuracy and real-time performance of small targets (e.g., tennis balls) in tennis scenarios. Aiming at the problems of loss of high-frequency information and redundancy of detection head parameters due to the traditional pooling operation, the researchers introduce a Wavelet Pooling (WP) module based on wavelet theory to preserve the high-frequency details through frequency domain decomposition, and design a lightweight shared convolutional detection head (LSCD) to reduce the number of parameters and improve the computational efficiency. Experimental results show that LWP-YOLO outperforms the YOLOv8 and YOLOv10 baseline models in terms of detection accuracy and speed, while significantly reducing the number of model parameters, making it suitable for edge device deployment.