Skip to main content

Fixations

A fixation is a camera's detection of a vehicle: a snapshot, a timestamp, the camera, the recognised registration plate and a recognition confidence.

Open the page from Monitor → Fixations or at /fixations. The section subtitle reads "Vehicle detections from cameras".

Filters

The filter bar at the top of the page:

FilterPurpose
Search plate or cameraText search across both columns at once
From / ToA time range accurate to the minute
All camerasPick one or several cameras; the counter shows "3 of 12 cameras"
ConfidenceAny confidence, ≥ 70%, ≥ 85%, ≥ 95%
Plate detectedShow only fixations where the plate could be read

The counter beside the heading shows "1,248 fixations" — how many records match the current filters.

With no cameras selected you get No cameras; with no matches, "No fixations match your filters" plus the hint "Try a wider time range or clearing filters."

View modes

The switch on the right chooses the presentation:

  • list view — compact rows: snapshot, time, camera, plate, confidence. Good for scanning large volumes.
  • grid view — cards with large snapshots. Good when the image itself matters.

Records are sorted newest first and load as you scroll ("Loading more…").

What a record contains

FieldMeaning
SnapshotThe frame at the moment of fixation
TimeDate and time
CameraWhich camera recorded it
License plateThe recognised plate, or Not detected
VehicleThe matched vehicle, if one was found
Conf.How confident the system is in the recognition

Records whose date could not be determined are grouped under Unknown.

Export

Export CSV downloads the current selection with all filters applied. The file columns are:

  • Date
  • Time
  • Camera
  • Plate
  • Country
  • Vehicle confidence
  • Plate confidence

Reading the confidence figure

Confidence is an estimate of recognition quality, not a guarantee. A practical rule of thumb:

  • ≥ 95% — usable without manual review;
  • 85–95% — usually correct, but check the snapshot before anything with legal weight;
  • 70–85% — needs a human to verify;
  • < 70% — usually noise: a dirty plate, a bad angle, glare.

Vehicle confidence (the object was detected and classified) and plate confidence (the plate was read) are reported separately. The first can be high while the second is low — the vehicle is visible but its plate is unreadable.

  • Cross-camera figures live on the Camera analytics screen, which has an All fixations block and a Detections by class chart.
  • A specific camera's live picture and archive are on its camera page.
  • The Camera fixations figure for a period is on the dashboard.

Typical workflows

Find a vehicle by plate. Type the plate into the search field, widen the range with From / To, and drop the confidence filter if needed — partially recognised plates may not match an exact query.

Prepare a report for one observation point. Select a single camera, set a 24-hour range, switch on Plate detected and press Export CSV.

Assess recognition quality. Set the filter to Any confidence, switch to grid view and look through the low-confidence snapshots — the cause is usually visible: angle, glare or a dirty plate.