Yu Chu¹ and Garret Washburn², ¹Nanchang University, China, ²California State Polytechnic University, USA

Basketball is one of the most popular sports worldwide, with over 610 million people aged 6 to 54 playing the game at least twice a month, according to FIBA. However, access to systematic and professional basketball training remains limited, especially in developing countries, where only 1%–3% of players may receive professional coaching. This lack of access makes it difficult for most basketball enthusiasts to learn and refine proper shooting techniques. To address this issue, we propose Sharp Shooter—a mobile application that helps users improve their shooting form without requiring professional training or expensive equipment. Our solution combines several cutting-edge technologies: MediaPipe is used to extract key body landmarks from users' uploaded shooting videos; these landmarks are then analyzed by a large language model (LLM) to provide expert-level feedback [10]. Additionally, the app matches users’ shooting forms with those of professional NBA players stored in a custom database, allowing users to see which NBA player their form most resembles— further enhancing engagement and motivation. The project integrates several key components, including a cross-platform front end built with Flutter, a Flask-based backend hosted on AWS, and a machine learning pipeline utilizing YOLOv5 for object detection and Random Forest or LSTM for motion quality assessment [11]. During development, we encountered challenges related to efficient video processing, backend scalability, data security, and precise motion evaluation. These were addressed through asynchronous data handling, load balancing, encrypted communications, and model optimization. The app was tested in various real-world use cases, including indoor and outdoor shooting scenarios, different lighting conditions, and varied camera positions. It consistently delivered actionable feedback, helping users recognize flaws in their form and track improvement over time. Our findings demonstrate that Sharp Shooter is a scalable, accessible, and affordable tool for basketball players at all levels. It offers a novel way for individuals—especially in under-resourced communities—to receive professional-style feedback and engage with the game in a more meaningful and data-driven way.

MediaPipe, Random Forest, Mobile Application, Shooting Motion, Large Language Model