2021/05/20 by William McNally, McNally, William, Pascale Walters +7 · 1 citation
Economics, Econometrics and Finance · Engineering · #Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Sports Analytics and Performance #Sports Dynamics and Biomechanics
paper · pdf · doi:10.48550/arxiv.2105.09880
openalex publication_date 2021/05/20 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Existing multi-camera solutions for automatic scorekeeping in steel-tip darts\nare very expensive and thus inaccessible to most players. Motivated to develop\na more accessible low-cost solution, we present a new approach to keypoint\ndetection and apply it to predict dart scores from a single image taken from\nany camera angle. This problem involves detecting multiple keypoints that may\nbe of the same class and positioned in close proximity to one another. The\nwidely adopted framework for regressing keypoints using heatmaps is not\nwell-suited for this task. To address this issue, we instead propose to model\nkeypoints as objects. We develop a deep convolutional neural network around\nthis idea and use it to predict dart locations and dartboard calibration points\nwithin an overall pipeline for automatic dart scoring, which we call DeepDarts.\nAdditionally, we propose several task-specific data augmentation strategies to\nimprove the generalization of our method. As a proof of concept, two datasets\ncomprising 16k images originating from two different dartboard setups were\nmanually collected and annotated to evaluate the system. In the primary dataset\ncontaining 15k images captured from a face-on view of the dartboard using a\nsmartphone, DeepDarts predicted the total score correctly in 94.7% of the test\nimages. In a second more challenging dataset containing limited training data\n(830 images) and various camera angles, we utilize transfer learning and\nextensive data augmentation to achieve a test accuracy of 84.0%. Because\nDeepDarts relies only on single images, it has the potential to be deployed on\nedge devices, giving anyone with a smartphone access to an automatic dart\nscoring system for steel-tip darts. The code and datasets are available.\n