A man who developed a system to lock the door for the cat 'only when the cat caught the prey'



The

cat , which is loved as a pet all over the world, has become close to humans since it was domesticated for the purpose of catching rats. It is a cat with a leisurely image, but sometimes it sometimes catches birds, insects, rats, etc., in memory of wild instincts. Ben Ham, who keeps the cat, was having trouble getting the prey caught by the cat to the room in the middle of the night, so she used machine learning to 'only when the prey was caught, the cat had a dedicated door It is said that we have developed a system that prevents you from entering the room.

Cats, Rats, AI, Oh My!-Ben Hamm-YouTube


The man standing at the right end of the stage is Mr. Ham.



And, this metric is Mr. Ham's love cat.



Although the metric is very cute, mice have begun to appear around the apartment where Mr. Ham lives ...



It is said that it has become a 'ruthless serial killer' that can not help catching rats from wild instincts. In addition, it is a dummy mouse prepared by Mr. Ham that the metric is added in this photo.



Metric, who turned into a serial killer, caught the rat at a rate of about once every 10 days, and visited Ham's room from the cat-only passage outside at 3 o'clock in the middle of the night. In most cases, although the mouse which Metric captured is already dead, sometimes the mouse is injured but not yet dead, in which case it was necessary for Mr. Ham to euthanize the mouse. Of.



Ham, who had been plagued by such days, came to think that he wanted to improve this situation somehow.



Possible measures include putting tins on the metric and putting a neckrest on it to make it impossible to hunt rats, or to shut it out completely in the middle of the night.



Although it is easiest to shut out the mouse to keep it from bringing in the rat, Metric visits the room not only when it catches the rat but also when it wants to use the cat litter box. Therefore, if you shut it out completely, you have to take care of your poo the next morning and it is troublesome.



So Ham came to want to develop a system to selectively lock the door only when Metric was catching a rat.



The outline of the system looks like this. Operate the lock mechanism using

Arduino on the microcomputer board.



The problem was how to lock the door only when Metric grabbed the mouse. The left is the metric that did not catch the rat, and the right is the metric when it caught the rat. The metrics that are converted to serial killers on the right have a sharp eye impression, and have an atmosphere that evokes wild instincts.



Mr. Ham decided to use machine learning in this process. In machine learning, teach a computer that 'Lama is such an animal' ...



It can be made to distinguish llama from various animals. Mr. Ham challenged to do this with 'normal time metrics' and 'metrics that have caught prey.'



Ham bought a camera on Amazon and installed it by the window, and started shooting the figure of the metric coming up the cat-only passage. It takes only two seconds for the metric to move through the path, during which time the system must determine if the metric has caught the rat.



The installed camera captures 23,000 images in a couple of months. Mr. Ham 'does not have a metric', 'the metric has not come to this side', 'the metric has come here but does not hold the mouse', 'the metric has come with the rat We did four kinds of tagging '. In all the images, it is said that 260 pieces of metrics had mice.



And Ham started using a tagged image to train a three-step machine learning model.



The first model allows you to determine if there is a metric, and then allows the second model to determine if the metric is coming to you.



Finally, in the third model, training was conducted to determine whether the metric is holding a mouse.



'If you have a mouse, you have three actions,' said Ham.



First of all, the cat passage is locked, and secondly, a message is sent to the smartphone. And third, donate to the

National Audubon Society of Nature and Environmental Protection. The last action seems to show the respect of Mr. Ham to the life of the mouse captured by the metric.



After starting the operation of the trained machine learning system, the metric climbs the passage about 180 times in 5 weeks without adding a mouse, but it is only 1 time accidentally shut it out. On the other hand, if you catch a mouse, it seems to have succeeded in locking with 80% accuracy.



By the way, this is an image taken from the inside of the room, the metric that has been locked out. You can see the metric puzzle, “Why not?”, Which says that Mr. Ham was able to get a peaceful night.



The end of the movie ended with a gratitude for the metric.



in Software,   Hardware,   Science,   Creature,   Video, Posted by log1h_ik