two honest ways to find a place
Most "geolocation" tools hand you a confident pin and hope you don't check. sleuth is the opposite: we tell you which signal we used, and how sure we are.
GPS in the file
Phones and some cameras write latitude and longitude into the photo's EXIF. We read those coordinates directly — precise to the meter, no model, no guessing.
a vision model reads the scene
No GPS? A model looks at architecture, vegetation, signage, road markings, driving side and light, and returns a country/region with a confidence score.
no signal? we say so
If neither path can answer, we tell you plainly. A wrong pin is worse than no pin — and we'd rather be trusted than impressive.
reading the GPS, locally
The EXIF block of a JPEG contains a GPS IFD with three rational numbers — degrees, minutes, seconds — plus a hemisphere reference. We parse it in your browser, before the photo ever leaves your device.
That pair of decimals is all a map needs. We pin it, then reverse-geocode it through OpenStreetMap's Nominatim to turn coordinates into a street, city, or landmark name.
letting a model look
When there's no GPS, we send a downscaled copy of the photo to a vision model on our edge. It's trained to notice the things geoguessers notice — and we show you its reasoning, not just its answer.
architecture
roof pitch, brick, balconies, building codes — regions build differently.
signage & language
what a street sign says, and what script it's in, narrows a country fast.
vegetation & climate
palms or pines, dry or humid, snow or dust — climate is a fingerprint.
road & driving side
markings, plates, and which side cars drive on cut the map in half.
light & shadow
sun angle and direction hint at latitude and time of day.
landmarks
a tower, a mountain ridge, a skyline — occasionally a dead giveaway.
try it on your own photo
the honest answer is usually the fastest one.