Cheng Chang¹ and Jonathan Sahagun², ¹BASIS Independent Silicon Valley, USA, ²California State Polytechnic University, USA

Accurate object detection is crucial in various computer vision applications. This paper explores the use of deep learning for the specific task of tennis ball detection in images, a task with applications in sports analytics and automated systems [1]. We propose a deep learning-based object detection model trained in a curated dataset of annotated tennis ball images. The model utilizes convolutional neural network architecture and is optimized using various loss functions. Challenges included data scarcity and variations in tennis ball appearance, which were addressed through careful data curation and potential data augmentation strategies [2]. Experiments were conducted to evaluate the model's performance in accurately locating and classifying tennis balls, with mAP as the primary metric. Results demonstrate promising performance, showcasing the potential of our approach. This focused approach, utilizing a tailored dataset and analysis of key performance indicators, provides a solid foundation for developing robust tennis ball detection systems, potentially impacting areas like automated officiating and player performance analysis

Robotics, Sonar distancing, Efficiency, Collection of tennis ball