The goal of technology is to make the world a better place. Sometimes, you can do that by making a gigantic breakthrough that solves a big problem in one fell swoop or opens major new horizons to the whole of humanity.
Much more often, however, you can solve a much smaller problem right at home. That’s what Makoto Koite, a cucumber farmer in Japan, did when he used Google’s open-source AI software TensorFlow, an Arduino, and a Raspberry Pi to automatically sort his produce by size and shape.
Google’s cloud platform blog has fuller details, but basically it works like this: the cucumbers get placed into a plastic chamber, and get their picture taken from the top, bottom and side. The picture then gets classified based on its similarity to several thousand pre-sorted images Makoto used for a training dataset. Finally, it’s dumped onto a conveyor belt, where a small motorized arm shunts it into one of several bins.
It’s not perfect yet, according to Makoto, but further developments in Google’s AI libraries and bigger sample sizes for training ought to make the cucumber sorter highly effective, letting the family concentrate on less mind-numbing tasks.
The Raspberry Pi doesn’t have to feel left out of the containerization gold rush anymore – the Raspberry Pi Foundation announced this week that Docker is coming to the Pi. Yes, it’s been out there unofficially for a while, but the foundation said that support for the revolutionary deployment framework is now built-in to the latest versions of Raspbian Jesse meant for the Pi.
What hobbyists can do with this, I’m frankly not sure – however, the foundation suggests the use of big groups of networked Pis for Docker’s Swarm mode, creating computing clusters. Cool. I guess if you’ve got a microservice lying around the house that you’ve been wanting to deploy, this is for you.
The Pi’s runaway success has spawned a ton of competitors eager to get into the market for small, cheap hobbyist computers. Two of the latest, the Omega2 and the PocketCHIP have intriguing upsides compared to the Raspberry Pi.
In the case of the Omega 2, that advantage is, surprisingly, cost – it costs just $5, but it comes with on-board Wi-Fi and solid-state storage. (There’s also a $9 version that doubles the $5 unit’s 64MB of memory and 128MB of storage.) After a successful Kickstarter, the first devices will ship in December.
PocketCHIP, in contrast, is somewhat more specialized where Raspberry Pi is dizzyingly generalist. It’s also quite cheap, at $9, and it packs some pretty impressive specs for the price, as the Inquirer notes in an engaging review. The point of the system, however, is gaming. It’s apparently designed to work with a (substantially more expensive) case mod that effectively turns it into a programmable Gameboy.
You can do much the same thing with a Pi, of course, but the price point, even with the $60 tacked on for the case, could be attractive to people looking specifically for a tiny, pocket portable gaming rig that requires a little less assembly.