An iOS engineer turned his weekly shooting‑range routine into a computer‑vision problem, aiming to replace the manual brass‑plug scoring used by NSRA clubs. He began with Apple’s Vision framework, hoping its object‑detection APIs would locate bullet holes on target cards. Early attempts mis‑identified ring markings and stray dots, forcing him to explore alternative preprocessing techniques.

He dug up a 2012 Warsaw University paper that described a four‑step pipeline—erase rings, flood‑fill, edge detection, then Hough‑transform circles. Implementing the first three stages with OpenCV succeeded, but the final Hough step also captured printed score digits. Tweaking the pipeline with a radial‑intensity profile raised detection to roughly 80 % accuracy on a mixed set of cards.

A later 2023 study showed YOLOv8 could pinpoint holes with 96.5 % precision, though it assumed perfectly aligned cards. Merging the geometry from the earlier OpenCV work with YOLOv8’s localization gave a robust solution; the model was exported to CoreML and packaged at 22.4 MB. The resulting app now scores NSRA cards on‑device without manual intervention.

The engineer plans to open‑source the code, hoping other shooting clubs can adopt the tool and reduce paperwork while preserving the sport’s traditional scoring ritual.