Edge AI

Custom analytics on professional cameras

Detection, classification and tracking adapted to unusual objects and the devices observing them.

A project from peoly’s published portfolio.

From input to useful output
  1. 01Camera video
  2. 02Detection & tracking
  3. 03Application events

The challenge: a specific object and behaviour

Some camera projects concern objects or activities that a general-purpose detector does not describe well. peoly’s published work includes light aircraft at small airports, urban waste bins, airport trolleys, cyclists and livestock.

Bringing the analysis to the device

peoly develops embedded software for non-standard object detection, classification, tracking and behaviour analysis. The portfolio covers static cameras and cameras mounted on vehicles. For more demanding workloads, additional embedded computers can take over part of the processing.

The published Axis partnership announcement adds concrete examples: tracking shopping carts in retail, analysing and streaming baseball games, and tracking light aircraft while reading their registration codes.

Matching the application to its surroundings

These projects share a practical requirement: the useful output depends on the target object, the view available to the camera and the action the receiving system needs to take.

A new deployment therefore starts with representative footage, the target device and a clear description of the required events or data. The published examples describe engineering experience; suitability for another device or scene is assessed for that project.

← Back to all projects

A similar challenge in your operation?

Tell us about your data, devices and the workflow you want to improve.

Discuss your project ↗