Description

In-pocket mobile classifier to detect when the user has performed a golf swing based on accelerometer readings. Records location of all swings via smartphone GPS, to enable playback of a round, including distances, and shot placement statistics. Requires no manual input, unlike comparable products on the market.

Highlights

  • Cross-platform event-based C++ code that runs on iPhone and Android
  • Robust signal processing and feature extraction
  • Machine learning data pipeline
  • Efficient code takes up a fraction of iPhone 5 CPU while detecting swings

Demo

Golf Swing Recognition Demo

Market Summary

The golf economy is $60 billion, annually, with golfers spending $6 billion on equipment. Products that give golfers statistics on their rounds are achieving some early success. For example, Game Golf has raised $8 million dollars, and GolfShot has made over $50 million in revenue. However, both applications require a substantial amount of input from the user. A product that requires no user input to track statistics, and that delivers meaningful advice could dominate the market.

Technologies Used

C++, Boost Asio, dlib Machine Learning Library, Bash and Python Machine Learning Pipeline, dygraph Javascript visualization library, Git