Visually detecting items in your home projects has generally been an arduous task. Pixy makes it very easy.
The task of visually tracking things is so complex and processor-hungry that it generally ends up taking over the entirety of a project's scope as opposed to being a piece in the puzzle. The hobbyist tends to stay away from this because many development boards like the Arduino may not have much horsepower left after analyzing visual data. Pixy changes that.
The idea behind Pixy, a Kickstarter project sponsored by Charmed Labs and Carnegie Mellon, is that you should be able to have visual recognition as easily as you have any other simple input device. They're offloading the processing to the Pixy board and only passing the minimum of information to the Arduino, freeing it up to do other things with that information. This should really open the doors to more hobbyists building robots or systems that have visual item recognition.
Pixy doesn't have to be used as an accessory to a microcontroller. It is actually quite capable on its own. You can connect directly to the board and use it to drive circuits via SPI, I2C, digital, and analog. However, connecting to a dev board might allow you more options, like using multiple Pixy boards for a wider field of vision.
Here are the specs listed from the Kickstarter page.
- Processor: NXP LPC4330, 204 MHz, dual core
- Image sensor: Omnivision OV9715, 1/4", 1280x800
- Lens field-of-view: 75 degrees horizontal, 47 degrees vertical
- Lens type: standard M12 (several different types available)
- Power consumption: 140 mA typical
- Power input: USB input (5V) or unregulated input (6V to 10V)
- RAM: 264 kbytes
- Flash: 1 Mbytes
- Available data outputs: UART serial, SPI, I2C, USB, digital, analog
- Dimensions: 2.1" x 1.75" x 1.4"
One major aspect that tends to limit tracking is the speed the camera samples at. Pixy captures at 50 frames per second, which you can see is fast enough to track many objects moving quite quickly. The tracking is done by color, and training it for a new object simply requires holding the object in front of Pixy and pressing a button. Once the object is "learned" it will begin tracking. While it won't do facial recognition, its developers say that it should be able to do facial tracking in the near future.
The Kickstarter for Pixy has two weeks left to go but has already far exceeded its goal. The backing levels at $39 and $49 are all sold out, but you can, as of the time of this article, still get one at $59.