The product of a partnership between Carnegie Mellon University and Austin-based Charmed Labs, Pixy is the latest version of the CMUcam, a popular line of open source vision sensors. The Kickstarter has already raised its goal of $25,000, and Pixy is well on its way to reaching its goal of providing an easy-to-use, smart camera sensor at low cost for a wide audience.
Until now, robotic vision systems with object recognition have been pretty exclusively applied in research and industrial grade robots?mainly because of the high demands on processing power necessary for crunching visual data. But thanks to Pixy, lighter-weight robots can ?see? and track objects without needing the big picture (and subsequent big data).
Pixy operates a bit like high-tech ?I Spy.? Essentially, you train the camera to identify objects by holding an example for detection in front of the lens while pressing a button. It?s that simple.
The sensor board generates a statistical model of the shades in the particular object. From there, Pixy finds similar color signatures using its dual-core processor capable of examining 50 frames per second. The camera can track up to seven colors at once, processing a 640×400 pixel image frame every 20 milliseconds, or the closest to real-time imaging non-industrial robots have ever been.
If seven colors isn?t enough, Pixy also can be programmed to recognize specific color codes, or multiples swatches grouped together. Using this system, the camera can plot and report the real-time location of hundreds of objects. Color coding holds promise particularly for robotics hobbyists, as it can be used as location identifier. Labeling doorways with a unique color code allows for a Pixy-equipped robot to navigate to a destination. Or, if you want your robot to charge itself, program a code and label for the outlet.
Pixy is able to report real-time data through a number of interfaces, including UART serial, SPI, I2C, and digital or analog I/O. The PixyMon app developed for PC and Mac allows you to view what the sensor is seeing, and also helps with debugging apps used in building a Pixy-compatible robot.
Currently, Pixy is easily compatible with the popular Arduino micro-controller (in fact, the Kickstarter offer includes an Arduino cable with the vision sensor board for $59), but as an open source sensor, a broader spectrum of integration is available.
“We tried to make Pixy as easy to use as possible. We think this will make it popular with the robotics and maker communities,” Anthony Rowe, CMU faculty member said.
“We’ve opened up the design by using the Open Source Hardware licensing model. You get source code, schematics, board layouts, everything,” said Rich LeGrand, Charmed Labs President. “We expect almost everyone to use Pixy as-is, but we also hope that by opening up the design, others will be able to easily build on Pixy for their application.?
