Last year, Google released an artificial intelligence kit aimed at makers, with two different flavors: Vision to recognize people and objections, and Voice to create a smart speaker. Now, Google is back with a new version to make it even easier to get started.
The main difference in this year’s (v1.1) kits is that they include some basic hardware, such as a Raspberry Pi and an SD card. While this might not be very useful to most Hackaday readers, who probably have a spare Pi (or 5) lying around, this is invaluable for novice makers or the educational market. These audiences now have access to an all-in-one solution to build projects and learn more about artificial intelligence.
We’ve previously seen toys, phones, and intercoms get upgrades with an AIY kit, but would love to see more! [Mike Rigsby] has used one in his robot dog project to detect when people are smiling. These updated kits are available at Target (Voice, Vision). If the kit is too expensive, our own [Inderpreet Singh] can show you how to build your own.
Via [BGR].
I wish Google didn’t over hype this, it is just some python code and a few neural network models. Less than 50 megabytes of data, the rest is just a computer with accessible I/O lines that runs Linux. You can do that right now with whatever gear you have laying around. Seriously do they really need to sell any of them at all, from a revenue perspective? I would be actually impressed if they gave away a few fully functional labs to needy schools in poor nations. Make that girl’s schools, in Pakistan etc, if you like collecting SJW badges.
The Vision kit includes a daughter board with a Movidius MA2450 vision processor. It does the heavy lifting, not the Pi.
I dislike the idea that simply improving the world is dismissed as just collecting SJW Badges,
Let’s take it easy. Sending free Pi’s to the Middle East is a tough sell for “improving the world”. There’s a lot of other stuff higher up their list of needs you could send if you’re looking to actually do something positive.
Reminds me of this:
https://www.youtube.com/watch?v=_u7GxxirXyc
Most of the code is probably nothing to write home about but the vision hat they call it a bonnet is potentially very useful.
Instead of suggesting to use it for something silly and stupid like detecting if people are smiling a more useful application would be allowing a robot to do edge detection for obstacle avoidance or a drone to avoid objects around it.
What I might use one for might be to add vision to a big track or a robotic lawn mower so it no longer runs into things yes I could use bumper switches and proximity sensors instead and for larger projects I would use those as well.
There are plans to let you eventually run your own tensorflow models. The MA2450 has around 100GFlops.
Normally you’d have to use something like an i7 or a beefy PPC to get that kind of number crunching.
I see something like this being incorporated into outdoor/indoor IP cameras because that’s where it makes the most sense.
I OBJECT to this article…..did the kit pick up my objection :-)
“Vision to recognize people and objections” – I object to this vision!
Error: Objection not recognized.
I’m mostly interested in the Movidius daughter board. Does anyone have any experience with this new kit ? And seeing now it costs basically the same as the Movidius compute stick, which would be recommended/is more flexible ?
“The new AIY Voice Kit and Vision Kit have arrived at Target Stores and Target.com (US) this month and we’re working to make them globally available through retailers worldwide. Sign up on our mailing list to be notified when our products become available.”
So they’re still unobtanium!
And by the time they’re actually available globally… they’ll announce the new successor product and restart the cycle again with no product available?
Our local Target stores has them on the shelf, near Boulder CO.
Picked up mine no problem
Having the MA2450 in a position where it’s essentially a “man in the middle” in the high-bandwidth CSI pipeline between the Pi camera and the Pi opens up a lot of potential.
It’s a lot better than going Camera –> CSI –> BCM2837 –> USB –> MA2450 –> USB –> BCM2837.
Other than the Neural Compute Stick, Movidius/Intel essentially *doesn’t exist* except for billion-dollar customers. No development boards, tools, documentation, datasheets, reference designs, no availability of any silicon, nothing.
That makes the AIY Vision Bonnet tempting, because there is no other similar product available anywhere at all.
Personally I think the cardboard origami thing is a bit of fiddly pointless crap.
Just make it smaller, integrate it onto the form factor of a Raspberry Pi Zero (which they’ve done) and maybe just have the microphone on the main PCB (do you really need two?) and the big illuminated button built into the PCB and everything all integrated in one, except the speaker on flying leads some distance away.
I think there must be someone who has a degree in designing card board stuff and it probably was what he did before working for Google.
Packaging design (yes an actual field).
http://www.packagingdigest.com/
The new voice kit has the microphones directly on the PCB. There is also a button on the board, so really no need for the remote button, unless you want to use the RGB leds. The new kit also uses a proper audio codec (ALC5645). Cardboard is just a convenient, inexpensive enclosure that allows user experimentation, since the expectation is that many users will use the components in their own projects.
Agreed. It looks like an interesting piece of silicon. However lack of documentation and tools makes it a black box.
I couldn’t get mine to boot :/
I imagine if it was an AI kit, then it would have actual AI chips.
Uh. The vision kit at least does.
They hide it well.
https://aiyprojects.withgoogle.com/vision/
That is some nice flat-cable and cardboard origami.
Don’t worry, it comes with a privacy LED.
40 / 90 USD at Target.
Target’s site says 49.99 for the Voice Kit. perhaps it varies by area.
This doesn’t familiarize kids with sending all their data to Google, no I’m sure it doesn’t.
Is it available in Europe?
Micro Center currently has the previous AIY Voice Kit for $4.99!
Do you think you could use this vision kit to detect when and what birds are at your bird feeder? My wife and I debate what birds come when, would be fun to get data from that. The birds are pretty distinct. Do you guys know of any DIY projects that do that?
There is support for the inaturalist model. Should be able to do what you want to do.
https://aiyprojects.withgoogle.com/model/nature-explorer
Perfect, thank you. Now I have to spend the $90 to see if it will work.
Pigeons, house sparrows and starlings we have positive ID, electrify the perch. Zap! Let’s make the birds of America great again.
While this might not be very useful to most Hackaday readers, who probably have a spare Pi (or 5) lying around
Get out of my house, Drew!
FWIW, I think it’s more like 8 or 9 Pis of various vintage, a few TI dev kits, half dozen ESP12Fs, some NRF12L01s, a couple of BBC:Micros, a SAM D board, an MPLAB kit, a pocket Beagle, a few teensies, and probably a few more dev kits I don’t recall.
I’m so looking forward to retirement…
I have a similar pile of devkits and unfinished projects, yet I feel compelled to head out right now on my lunch hour to snag one of these from Target.
That Movidius coprocessor is interesting, and… Oh is that an unpopulated USB Type C jack?
I can’t understand all this marketing hype around stupid marketing words for the masses… neural, big data, Ai and deep learning, blockchains, etc …are only marketing words, to sell same things as ever, nothing more.
YOU might want to keep an eye on the Humble Bundle website. They’ve recently had book bundles covering a lot of the topics introduced on HaD. Reasonable priced with tiers so you get what you want, with helping charities being a bonus.
I think the negative, knee-jerk anti-google comments are great.
Yes. Google is Skynet, deal with it. The world in which we live, greater evils and all that jazz.
Stop and take a look at what it *is*. It’s some very esoteric hardware that a couple big *digital* companies are coming together and getting into brick-and-mortar stores, with profit as a minimal concern (keep in mind- when Voice was new MicroCenter was selling them for $5? if you bought a Pi3 at the same time. I bought 3.
It’s a complete kit. In a box. When I built my Voice kit, I think the only things I added were a bit of scotch tape and some hot glue (only because I wanted to make hot-glue-feet to keep it from sliding around). Think I took the tape off the mic board and added glue there too, since the gun was hot and all. Point is… I don’t have kids, but could very easily see kits like this being a rainy Saturday project…
Cardboard is great because it keeps the price of the kit down, gives it a different feel (I’m one of those people, man… I get a kick out of talking to a cardboard box. We had a hell of a conversation one night that had me looking at it from the corner of my eye like ‘wth skynet?’) and things like… It’s a good enough case if you don’t have a 3D printer, parts of it are easily replicated with on-hand materials if a fold gets mangled during assembly…
It’s supposed to be somewhat community driven, something in the spirit of a corporate backed OSS/OSH. The starter software is meant to be basic- primary build is Google saying ‘here… build this thing that is a lot like this other thing that we *do* sell expressly for the intent of profit.’ And the thing with neural nets? they are incredibly specific to what someone intends to do with it… they provided some basic ‘is this an apple? Is this an orange?’ models so you can demo it, geek out, and get started researching your own path- all in one day.
This is like the free-PiZero/MagPi thing… because they didn’t print an extra 7 million copies that would only be sold for this one-off promotion, or people didn’t live close to a brick-and-mortar, or Pi Foundation didn’t wait until they had 17 million square feet of warehoused Pis just waiting to ship so they could satisfy *all* the demand, *all* at once… two things Pi Foundation have proven is their ability to make hardware revisions within a core product line (ie: the new 3+ or the how many versions of the Zero, now?) and their ability to play against production and development time. Google/Pi played this one smart-ish and were able to mitigate the larger issues before commiting resources to something with hidden flaws. The early limited releases were essentially public beta testing. By not over publicizing it and limiting purchases to in-store people were essentially ‘opting-in’ to the beta… I don’t ever feel they were misleading about the state of the kits and accepted no-questions returns of defective kits. …which MicroCenter is typically cool about anyway.
So. That said… I’ll continue patiently waiting until MicroCenter starts getting Vision kits because that is my one complaint and something I’ll keep struggling with because as much as I want a Vision Kit, I’d prefer to give my support to MicroCenter. But maybe I’ll go check my local Target and buy four more when MC gets theirs… principles suck sometimes ;D
A fully open API and tool kit can eliminate the problem of giving your data to google.
The whole idea of chips like the Movidius is to allow a smart device to have it’s intelligence internally vs having most of it on some remote server you don’t control.
Based on Rasberry pi? ????????????
As much as technology births a lot of opportunities , it also poses immense threat to the human race if not curbed properly…. I see technology as a double edge sword…its power and ability lies in hands of its creators.
This whole unveiling is quite interesting!