Using AI to Score Split the G

The blog post discusses how an AI-driven web application, inspired by a popular Guinness-drinking trend called "Split the G," was developed to analyze and score the perfect pour. This trend involves drinking Guinness so that the line between liquid and foam lands halfway through the "G" on the glass. To accurately score the splits, the creators built two AI models: one to detect the beer, glass, and logo, and another to refine the detection of the "G" logo, achieving high precision in identifying a split and the beer level. They implemented a two-stage scoring approach, assigning scores from 0.00 to 5.00 based on the split's precision, and built a web app using tools like React Router, TailwindCSS, and Supabase for database management. The application captures images with Roboflow's InferenceJS to detect splits in real-time and uses the Roboflow Workflows API for inference. The app also features a leaderboard to encourage competition, displaying the top 15 scores and allowing users to share their results.