Introduction
GeoInfer is an image geolocation API. Send a photo and get ranked location predictions with GPS coordinates, estimated from the pixels alone. No GPS, EXIF or metadata required.
GeoInfer estimates where a photo was taken from what is visible in it. You send an image, and you get back ranked candidate locations with coordinates, place names and a confidence radius, usually in under a second.
It is one trained model that reads the pixels: architecture, terrain, vegetation, road layout and infrastructure. It does not search the web, run a reverse image search, read the EXIF header or call agents. That is why it works on screenshots, re-uploads and photos with the metadata stripped.
How accurate is it
Measured on the IM2GPS3k benchmark, using pixels only:
| Model | Within 1 km | Within 25 km | Within 200 km | Correct country |
|---|---|---|---|---|
| Global | 50% | 61.20% | 78.93% | 93.3% |
| High accuracy (regional) | 97% |
High-accuracy models cover specific regions. See models for what each one returns.
How it works
- Send an image to
POST /v1/prediction/predictwith your API key - The model analyses the scene
- You get ranked locations with coordinates, place names and a radius in km
Ways to use GeoInfer
REST API
First call in 2 minutes with curl, Python or JavaScript.
Python SDK
pip install geoinfer. Sync and async clients.
Telegram bot
Send a photo to @geoinferbot and get the location back in the chat.
Web app
Upload images, see results on a map and keep a history.
Chrome extension
Right-click any image on the web to locate it.
Maltego
Geolocate image URLs inside your investigation graph.
What people build with it
- OSINT and investigative workflows
- Verification of news, social media and user-submitted photos
- Insurance claim and property photo checks
- Batch geolocation pipelines over large image sets