Living in Ho Chi Minh City, I often travel to work by motorbike. Sometimes, I don’t realize it is raining on my route until I am already on the road. That is why I built RainML as a fun experiment to check for rain in real-time.
You can see the live app at rainml.pixelcarrot.com. It uses public traffic webcams, Uber’s H3 hexagon grid, and machine learning to make a live rain map of the city.
How It Works
The system regularly gets camera images, checks for rain, and groups them:
- Find Cameras: Get all active traffic cameras within 40 km of the HCMC center using Windy’s API (about 636 cameras).
- Hexagon Grouping: We put the cameras into Uber H3 hexagons (resolution 7). We only pick one camera per hexagon to keep it fast.
- Rain Check: We check the camera images using machine learning to see if it is dry or raining.
- Update Map: The rain results are visualized on the map.
Note: Right now, the machine learning model works well for detecting rain during the day, but it often gives false detections at night.
More details on the machine learning pipeline and H3 spatial resolution to be written here.