Google AIY Projects brings do-it-yourself artificial intelligence to students and makers. The AIY Vision Kit from Google is an intelligent camera that can recognize objects, detect faces, and emotions. Download and use a variety of image recognition neural networks to customize the Vision Kit for your own creation. What is included in the box: Raspberry Pi Zero WH, Pi Camera V2, Micro SD Card, Micro USB Cable, Push Button
STEM toys are Science, Technology, Engineering and Math based products that are geared towards developing useful skills in all areas of learning. This product focuses on the core subject: science.
STEM toys are Science, Technology, Engineering and Math based products that are geared towards developing useful skills in all areas of learning. This product focuses on the core subject: Technology.
STEM toys are Science, Technology, Engineering and Math based products that are geared towards developing useful skills in all areas of learning. This product focuses on the core subject: Engineering.
STEM toys are Science, Technology, Engineering and Math based products that are geared towards developing useful skills in all areas of learning. This product focuses on the core subject: Math.
Just an upfront, I am a very technical person, electrical engineer. I cannot for the life of me figure out how to get HDMI working. I've looked through help online, changed the config.txt, etc. Nothing works with HDMI.
Tried 3 diff HDMI monitors and two cables.
Good for DIY makers wanting to incorporate vision AI into your own projects
4 out of 5 stars
David Builder - 1 year ago, Verified purchaser
This is a nice product to experiment with inference models for your own electronics projects. It works fairly well but does occasionally freeze/lock-up under heavy use. Intel sells the computation device as a stand-alone product (Movidius Neural Compute Stick) for around the same price as this kit, but with this you get a pi, camera and other parts for around the same price so it's a pretty good deal. This is best for those DIY types wanting to incorporate vision AI into your own projects since it doesn't require an internet connection to perform inferences. There's a nice case you can print from thingiverse (attached photo). I removed one star since there's not a lot of guidance from Google on how to use your own models besides the ones included in the bundled apps.
DisappointedAIGuy - 1 year ago, Verified purchaser
Not recommened. Not much to say other than that. The inference perfromance is really bad: slow fps and terrible accuracy. Latest code and the optional models are just as bad. For those that think this is great then you have no idea how CNNs should work. Google should really look at this product and either redo it or pull it from sales as it will only damage their reputation over time.
Not for Novices, might be faulty, but Great starting point for Computer Vision Projects
4 out of 5 stars
AIYFan - 1 year ago, Verified purchaser
I'll preface this with some summary statements: Recommended prerequisites: some knowledge on Linux, Raspberry Pi & Programming. Additional adapters to use the kit with a Monitor, Keyboard & Mouse as listed on their instruction website. Warning: might be faulty, but there are fixes on their Github repository. Let's just get the negatives out of the way first. As some have pointed out, it seems v1.1 has some QC issues. I got 2 kits, 1 worked perfectly fine after building, the other will freeze/crash while running some of the demos. There are some proposed fixes if you check out the issues in their Github repository. I managed to fix my faulty one by changing some of the boot settings, though I am not sure of the long term consequences of the fix. Some knowledge on Linux, Raspberry Pi and programming is definitely needed, especially if you're one of the unlucky ones to get a faulty kit. Now onto the positives. The kit is definitely fun to build, though may not be for children and beginners as it requires some gentle handling of really small parts. But once the kit is complete (and hopefully it works), it is very exciting to play around with some of their demos and pre-trained neural network models. I would definitely recommend following the "use with monitor, keyboard & mouse" option as you will want a monitor to display the video feed of the camera. Programming the kit to do more with existing models should be pretty straightforward and fun as the source codes are all in the SD Card and on their Github. Creating custom models however, will require some knowledge of Deep Learning & Tensorflow. This kit will be a good starting point to learn about the hardware, software & Deep Learning components of Computer Vision. Happy hacking!
The Vision Kit has promise, but there appears to be a quality control issue. I got one that would crash when running any of the video related applications. The Raspberry Pi W board that is supplied runs at up to 1 GHz speed, but if it is allowed to run at full speed it crashes with a Linux kernel oops after a video related application has been running for a short period of time. I was able to improve this by changing the configuration to set the speed to 900 MHz. Google is aware that a number of people are receiving kits that have this problem, and others are not having the issue. I would give more stars but with this problem I had to spend a great deal of time isolating the issue and while I have a partial solution I expect that it will fail at some point in the future. If it's the Raspberry Pi W that's not a huge issue since that is a $10 part. If it's the vision bonnet that Google designed that is the issue then that is a bigger issue as I'm not sure that it is available separately. Google thinks that it is likely to be the Raspberry Pi W but there is no definite determination as of a day or two ago.
Easy to assemble, everything (except the USB power adapter) is in the box - couldn't be more convenient. Demos all worked immediately and are very compelling. Now I'm learning to train my own vision models. Very cool.