organizations in queue

Every pixel knowswhere it’s been.

Geographic anchors
110K+
TR-dense coordinate gallery
Street imagery indexed
9,488
Turkey · own retrieval index
Inference latency
< 1sec
Single-frame, top-K prediction
Embedding dimensions
512D
Visual · face · plate signatures

Mission

Every image is a witness. We teach machines to read it — turning a single frame into a place, in seconds.

Approach

We don’t search a database. We teach machines to see.

Older tools match a photo against a library of known places — and fail the moment the place isn’t in it.

Corvax reads the image itself: the architecture, the terrain, the vegetation, the light, the line of a rooftop. It learns how the world looks, so it can place an image it has never seen before.

System 01 — Tazı

Tazı reads the world from a single frame.

Our first system. Give Tazı one image and it returns where it was taken — ranked, with confidence, in seconds. No metadata required. The raven sees from above; Tazı tracks the trail.

01

geolocation

From image to region. Rank the most likely locations for a photo from its visual content alone.

02

street-level

From region to address. Narrow broad areas toward exact locations, even without an obvious landmark.

03

scene reading

From pixels to signals. Read architecture, terrain and context the way an analyst would — only faster.

Available to verified teams only

System 01 · in motion

Give it a frame. Get a place.

Input · single frame
SOURCE3024 × 4032 · JPEG
Prediction · top-1
SAKARYA · TR40.786, 30.406

No metadata. No manual search. One inference pass.

Beyond Tazı

Corvax is the home for a family of visual-intelligence systems.

Tazı is the first. It’s in private development now — more systems are on the way.

Access

Work with us.

Corvax is being built for verified teams and investigative work. Request early access and we’ll be in touch.

In queue
Verified & active