Technology

Attention is a pattern, not a point.

No single moment says what a driver is doing. AutoSight follows where attention goes and how long it stays away, and speaks up only when it matters.

Illustration · real face points

What it watches

It watches the eyes, not the phone.

Everything runs on the device, in AutoSight One or on the phone in the app, and no video is ever recorded.

An infrared view of a driver at night, the face evenly lit and the eyes clear.
Camera

Eyes

Where the driver is looking, not just where the head points.

Camera

Head

Which way the driver is facing.

Camera

Eyelids

Long blinks and heavy eyelids, for drowsiness.

Every moment

Time

How long attention stays away, and how often it goes.

Illustration · real face points

From glance to habit

Five steps, from a glance to a better habit.

  1. STEP 1

    Watch

    Many times a second, AutoSight checks where the driver's eyes are. Each picture is used on the device and thrown away.

    eyesheadeyelids
  2. STEP 2

    Time

    Every look away from the road gets a clock. Quick mirror and instrument checks are part of driving and never set off the chime.

    every glancemirrors are fine
  3. STEP 3

    Speak up

    At two seconds off the road, one short chime. Nothing to tap, nothing to read.

    chimehands-free
  4. STEP 4

    Review

    After the drive: a Focus Score, each drift moment, and Moment Replay, an animated face that shows where attention went. Never video.

    Focus Scoredrift momentsMoment Replay
  5. STEP 5

    Share

    If the family turns it on, each drive reaches a parent or a driving instructor, end-to-end encrypted, and the driver sees the same summary.

    familyinstructorencrypted

Temporal analysis

Three seconds tell a story one frame can't.

Good drivers look away from the windshield constantly: mirrors, blind spots, intersections, instruments. AutoSight treats that as driving. What it watches is where attention went, how long it stayed, and what came before and after.

In NHTSA's 100-Car study, off-road glances adding up to more than two seconds at least doubled crash and near-crash risk.[1] That two-second line is where AutoSight speaks up.

  1. 0.0 sRoadDriving.
  2. 0.3 sMirrorMirror check. No response.
  3. 0.6 sRoadBack on the road.
  4. 0.9 sDownGlance down. The clock starts.
  5. 1.9 sDownStill down after a second.
  6. 2.9 sDownTwo seconds off the road. Alert.

Hard cases

Built for the situations that matter most.

Sunglasses and eyewear

Eyes hidden? It says so.

AutoSight checks during setup whether it can see the driver's eyes. When sunglasses or glare hide them, it tells you how sure it is instead of guessing.

A phone below the camera

Read the driver, not the device.

Most phones in a car sit below any camera's view, in a lap or a cupholder. AutoSight follows where attention goes, so it doesn't need to see the phone.

Natural scanning

Mirror checks aren't distraction.

Good drivers look away from the windshield all the time. Quick checks never set off an alert, so alerts stay rare and meaningful.

Gaming the system

Looking compliant isn't enough.

A face held toward the road while the eyes keep dropping doesn't pass as attention.

What it tells you

It says what it saw, how sure it is, and when it can't see.

A beep isn't an explanation. Every drift moment in a trip summary says which kind it was, and how sure AutoSight is.

Eyes off the road

Attention left the road with no phone in view: a phone in the lap or console, a passenger, the floor, the center screen. Distraction isn't only a phone in a hand.

Probably a phone

When what it sees points to a phone, it says so, and says "probably," because that's what the evidence supports.

Quick glance

A mirror or instrument check under a second. Normal scanning, never an alert.

Eyes closed or drowsy

Eyes shut, or closing too often. A slower chime suggests a break.

Lower confidence

Sunglasses or glare hide the eyes, and AutoSight says so, instead of inventing where the eyes point.

Can't see clearly

The face is out of view. AutoSight says it can't tell, rather than guessing a cause, and the trip shows how long tracking was lost.

How we test

Results come with their method attached.

A single accuracy percentage hides who was tested and how. AutoSight's report shows its work.

A guided test, parked

The app tells the tester where to look, so every result is checked against what really happened. Nobody looks away on purpose while driving.

Counts, not slogans

Every result is shown as a count, like 9 of 10, with a 95% range. With a handful of testers, the range is wide, and the report says so.

Broken down

By glance type, by condition (glasses, sunglasses, night, infrared camera), and by what AutoSight called each glance, so weak spots can't hide in an average.

What it doesn't show

Real-road driving, eyes on the road with the mind elsewhere, and long-term drowsiness aren't measured by this test. The report lists that too.

Where it runs

One model, built to move between platforms.

PlatformSensorsUse
AutoSight OneInfrared camera on the dash with its own 940 nm lightNew drivers and families, day and night. The product we’re building now.
AutoSight appPhone front camera, motion sensorsTrip summaries and the family view, plus a phone-only mode to try AutoSight today.
FleetPhone or dedicated cameraAttention coaching and analytics for commercial drivers.
SDK and vehicle integrationExisting in-cabin camerasThe attention model as a layer inside other systems.
AutoSight One on a light studio surface: a slim black bar with a glossy front, a lens in the middle, two small infrared lights on each side, the AutoSight name at the lower right, on a short stand.
AutoSight One. Design concept; final product may vary.

Work on the hard parts with us.

If you work in computer vision, human factors, or driver safety, we want your critique.