This is a golf hole recognition program that demarcates various lies (e.g. fairway, rough, trees, etc.) on a satellite image, uses those demarcations to convert satellite golf holes to a visually enhanced rendering, and supports distance calculations between the tee box or hole to any point on the map.
Currently, there are no affordable APIs for satellite golf course mapping (e.g. the industry-leader, Golfbert, charges $399/month). This is prohibitive to developing satellite-based data entry for golf apps. Without satellite maps, data entry takes the form of a spreadsheet, which is inconvenient and time-consuming. This adds unnecessary friction for users and prevents wider adoption.
- Lies should be correctly identified. The full list of lies is: tee box, fairway, rough, super rough/fescue, trees, water, bunker, and green.
- Boundaries should be correct to 4 yards (e.g. the learned boundary between fariway and rough can be 4 yards right of where it actually is). For reference, a typical fairway is 40 yards wide.
- Boundaries should look smooth, rounded, and realistic (i.e. no jagged protrusions).
- The final rendering should be standardized and aesthetically pleasing (e.g. making all fairways a lime green to mask dead/brown spots on fairways). They should look roughly like the examples shown here: https://golfbert.com/api.
- The final rendering should be oriented upward (i.e. green should appear at the top of the screen, and tee boxes should appear at the bottom of the screen).
- Distance calculations should be accurate to a yard (e.g. if you click on a point in the fairway, the distance calculated from the tee box can be long/short by 1 yard).
- Tee box location data should be pre-loaded (e.g. if a user selects the blue tees, all distance measurements to points in the fairway should originate from the corresponding tee box).