Cropping and Zoom

Cropping works like a digital zoom that enlarges a selected area of your video before sending it to the cloud for analysis. Larger objects in the frame mean better AI detection - object size matters significantly for accurate recognition.

Why use cropping?

When objects appear small in your video frame (like vehicles on the far side of a roundabout or distant pedestrians), AI struggles to detect them accurately. Cropping solves this by:

  • Making small objects visible - enlarging distant or small objects so AI can detect them reliably

  • Improving classification accuracy - bigger objects provide clearer visual details for the AI to classify correctly

  • Focusing on what matters - removing irrelevant areas to concentrate processing power on your measurement zone

When to use cropping

Cropping is most effective when:

  • Objects are small relative to the frame (large intersections, roundabouts, distant areas)

  • You need to focus on a specific distant area (far pedestrian crossing, opposite side of intersection)

  • The measurement area is only part of the video (one lane out of many, specific entry/exit point)

How to set up cropping

  1. Open the Design tab in AITracker GUI

  2. Add or select a configuration

  3. Enable the "Crop" option

  4. Draw and adjust the rectangle:

    • Click and drag to position the cropping frame

    • Grab the handles to resize it

    • Use different aspect ratio from the dropdown for optimal AI input (recommended)

  5. The cropped area will be enlarged to full frame size during cloud processing

Important: Configure cropping before sending the video to the cloud - it cannot be changed after processing starts.

Working with multiple configurations

You can create several measurement configurations from a single video:

  • Configuration 1: Full frame for main intersection

  • Configuration 2: Cropped and zoomed on distant pedestrian crossing

  • Configuration 3: Cropped on far lane for better vehicle detection

Note: Each configuration counts as a separate cloud processing job and incurs additional costs.

Last updated

Was this helpful?