Khoj
खोजthe trace · the search — Footprint, Track & Movement Detection
What if a stretch of border no patrol can watch could tell you the moment a person — not an animal — crossed it?
Hundreds of kilometres of jungle border can't be fenced or patrolled end to end, and there's no cloud out there to send footage to.
The situation
An unwatched forest border at night
A cheap camera-and-sensor post sits on a smuggling trail deep in the jungle, off every network. Khoj, running on the post itself, sees a track form in the mud — and knows in a second whether it was a deer or a man, and which way he went. The section gets the one alert that matters, not a thousand that don't.
How it plays out
Step by step
- 01
Something crosses
A camera, infrared or seismic sensor on the trail picks up a track or movement.
- 02
Khoj reads the track
On the post itself, it classifies what made it — boot, paw or tyre — and the direction of travel.
- 03
It knows its own ground
It has learned the animals and patterns of this trail, so a deer doesn't trip the alarm — a person does.
- 04
One alert that counts
The patrol gets a single, confident warning of human movement, with no footage ever leaving the post.
The system itself
Under the bonnet
Reads imagery and unattended-sensor feeds along trails, borders and forest edges, and classifies what made a track — boot, paw or tyre — and which way it moved. It learns a patrol's own ground and flags fresh human movement through terrain no one can watch by eye.
What it means for you
The bottom line
- ▸Tells a human track from an animal one
- ▸Works offline on cheap edge sensors
- ▸Cuts false alarms by learning the local ground
- ▸Early warning where a fence and a patrol can't reach