|Machine Learning For Cat Control|
|Written by Lucy Black|
|Saturday, 06 July 2019|
This may not be the cat video you were expecting, but after numerous articles on computer vision recognizing cats, here's one in which machine learning is put to work to detect when your cat is bringing home a trophy you really don't want indoors.
Cats are natural hunters and they also like to show off their prowess, offering up the catch of the night for their owners to admire. While Ben Hamm, a product manager at Amazon, originally acquired his cat called Metric to deal with a rat infestation in his apartment, he really didn't want to have to deal with the dead, or worse still wounded but still alive, "presents" that Metric brought home in the early hours on a regular basis.
Hamm's solution for stopping his "sweet, murderous cat” from bringing home dead and half-dead prey in the middle of the night and waking him up was to use AI to create a smart cat flap that would prevent Metric from entering when carrying prey while still allowing him to come and go as he pleased otherwise.
The locking mechanism contraption was controlled by an Arduino and the system relied on an Amazon DeepLens camera and used Amazon SageMaker for the learning models. Having been advised to learn to code when he asked to hire a someone to help with the machine learning, Hamm hand-labelled a set of more than 23,000 photos to train the model.
The model had three stages. The first distinguihsed between No Cat and Cat; the second between Cat Going and Cat Coming and the third Cat Alone and Cat with Prey. The "Cat with Prey" state triggered the cat blocker, locking the cat flap for 15 minutes, texing images to Hamm and paying out "blood money" - a donation to the National Audubon Society, which protects the birds that cats like to kill.
Hamm gave the following 5-minute presentation at Ignite, Seattle which the audience greeted with lots of laughter as well as admiration for a great practical implemention of computer vision and supervised learning.
The system has proved its worth. Over a period of 5 weeks Metric has come through the cat flap "innocently" 180 times and has been unfairly locked out one. He has attempted to bring six dead trophies inside and has been stopped on all but one occasion. Ben Hamm has been able to sleep through the night and has had the added benefit of having learned to code.
Fun Kaggle Challenge To Tell Dogs From Cats
Smart Cat Feeder Uses Facial Recognition
To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin.
or email your comment to: firstname.lastname@example.org
|Last Updated ( Saturday, 06 July 2019 )|