Cat detector: a deep learning experiment


ANPR is based on deep learning, a machine is taught how to recognize number plates. AVUTEC?s Gatekeeper is an embedded ANPR camera, but also a versatile IoT camera. Therefore, it can do a lot more than identify vehicles. As it turns out, the Gatekeeper can also be used as a cat detector.

A deep learning model to detect cats

AVUTEC specializes in video analyzing techniques. This involves not only ANPR, it covers the recognition of anything, from people to all possible objects. With this in mind, AVUTEC decided to incorporate a deep learning model that detects cars, bicycles, trucks, people, dogs, birds and cats into computer vision platform CortexFramework.

As a deep learning experiment, the Gatekeeper was used as a cat detector to stop neighboring cats from using the planter boxes on AVUTEC?s terrace as litter boxes.

A construction with a garden hose  

The Gatekeeper was placed behind a window, where it had a good overview of the planter boxes in question. Then a construction was made with a:

  • washing machine hose
  • magnetic valve
  • garden hose with a sprinkler
  • power cable
  • Gatekeeper 410
  • I/O extender of the Gatekeeper

The sprinkler was placed in a planter box. As soon as the Gatekeeper would detect a cat, the I/O extender should trigger the sprinkler to discourage the animal.

The deep learning experiment

On a sunny day everything was set in place, but it took quite some time for a cat to show up. A piece of chicken on the edge of a planter box eventually helped and soon enough the Gatekeeper could detect neighboring cats. Of course, the sprinkler was activated every time, which resulted in some very funny footage.

Do you want to see how the cat detector works and how cats are chased away with a sprinkler? You can watch the video on our Youtube Channel: