Derek Li¹, Austin Amakye Ansah², ¹USA, ²California State Polytechnic University, USA

Traditional basketball training relies on feedback from coaches and video playback, which can be delayed and inconvenient for solo practice. This paper introduces HoopLab, a cross-platform mobile application designed to provide immediate, comprehensive analysis of basketball shots. The app, built with Flutter, utilizes a custom-trained YOLO object detection model to analyze user-submitted videos. It features two distinct analysis modes: a backboard view that confirms if a shot is successful by tracking the ball's path through the rim, and a side-view that evaluates a shot's quality by comparing its actual path to a computationally generated optimal trajectory. The underlying YOLOv11 model was trained on a dataset of over 8,200 images, achieving a mean average precision (mAP@0.5) of 94.72% and demonstrating high performance in identifying key objects like the ball, rim, and player[10]. HoopLab offers players a powerful tool for instant, data-driven feedback to accelerate skill improvement.

Computer vision, AI, mobile, basketball, Flutter