Shooting squirrels with a water cannon might not seem to be a serious project, but it involves AI and a lot of hardware.
If you compare what you can buy in terms of intelligent devices to what you could create with the latest technology, you have to notice that there is a big gap.
It is very similar to the early days of the home computer revolution. Back then there were pocket calculators or expensive minicomputers. Hobbyists brought the low cost machines that created a revolution to the market. First by building their own and later by selling their creations. We seem to be in a similar situation with AI and robotics at the moment and DIY projects seem to be the only way to get what you really want at a reasonable price.
Take the problem of a squirrel scarer. If you want to get rid of pesky squirrels, or any similar small creature, your only options are to get a gun or some low tech device that couples an IR sensor with a hose pipe and valve.
Kurt Grandis took some cutting edge and open source AI tools, Python, an Arduino and a SuperSoaker and built the perfect squirrel hosing machine.
To quote from his PyCon 2012 talk:
"Has your garden been ravaged by the marauding squirrel hordes? Has your bird feeder been pillaged? Tired of shaking your fist at the neighbor children? Learn how to use Python to tap into computer vision libraries and build an automated sentry water cannon capable of soaking intruders."
The project involved Open Computer Vision (OpenCV), an a SVM learning procedure that he trained to tell the difference between a squirrel and a non-squirrel. Some manual feature extraction was used in the initial processing. Squirelness seems to come down to "blob size", color and texture.
After "perfecting" the classifier the hardware came next - a SuperSoaker Mark I was used as the "water cannon". A pair of servos were used to aim the gun and a third to pull the trigger.
You can see the entire talk in the video below - but if you just want to see the squirrels get washed it happens at 16 minutes in.
The first version just didn't have the fire power. Squirrels seem to be so persistent that a SuperSoaker runs out of water all too quickly. The next version will be more sophisticated and have a much bigger water gun.
A new course on TensorFlow starts this month and sets out to introduce you to the state-of-the-art approach to building artificial intelligence algorithms. However, as it is on Kadenze, you can e [ ... ]
The Raspberry Pi Compute Module is aimed at the "professional" user and is a traditional embedded system, but it is looking a little low on specification. Now we have the news that it is going to be u [ ... ]