Sign in
Home / AI models and agents / People analytics / Forklift-pedestrian proximity
Verified Model People analytics

Forklift-pedestrian proximity

Sarvavyapi Vision, Pune. Version 2.3. Runs on EdgeBox Your servers.

Alerts when a person walks within 2 m of a moving forklift. Emits both boxes and a clip.

IndoAI test report Run on the people analytics footage set, 2026
Reliable for indoor warehouses in daylight and IR. Not recommended outdoors in rain or below 3 m. Edge latency under 1 s.
94%
catch rate
0.4
false alarms per camera-day
0.7 s
time to alert
91%
still active after 90 days

What it needs from your camera

Camera type
Bullet or dome, fixed lens
Minimum resolution
2 MP at 15 fps
Pixels per metre
40 or more at the far edge of the zone
Mounting height
3 m to 6 m
Lighting
Daylight or IR
Compute
EdgeBox Your servers

Data handling

Category
Person present. No identity inferred.
Frames stored
Only the 6-second clip around an event, on your EdgeBox or NVR.
Leaves the site
Events and clips you choose to forward.

Notifications

[wc_notifications]