DeepDarts: Modeling Keypoints as Objects for Automatic Scorekeeping in Darts using a Single Camera

Abstract

Transform 2. Calibrate Board 3. Score input image score prediction keypoint detections T20 1 18 Figure 1: Overview of DeepDarts, a method for predicting dart scores from a single image captured from any camera angle. DeepDarts uses a new deep learning-based keypoint detector that models keypoints as objects to simultaneously detect and localize the dart coordinates (purple) and four dartboard calibration points (yellow). The calibration points are used to transform the dart locations to the dartboard plane and calibrate the scoring area. The dart scores are then classified based on their position relative to the center of the dartboard.