Introducing Google Coral: Building On-Device AI (Google I/O'19)
Vložit
- čas přidán 30. 05. 2024
- This session will introduce you to Google Coral, a new platform for on-device AI application development and showcase it's machine learning acceleration power with TensorFlow demos. Coral offers the tools to bring private, fast, and efficient neural network acceleration right onto your device and enables you to grow ideas of AI application from prototype to production. You will also learn the technical specs of Edge TPU hardware and software tools, as well as application development process.
Watch more #io19 here: Machine Learning at Google I/O 2019 Playlist → goo.gle/2URpjol
TensorFlow at Google I/O 2019 Playlist → bit.ly/2GW7ZJM
Google I/O 2019 All Sessions Playlist → goo.gle/io19allsessions
Learn more on the I/O Website → google.com/io
Subscribe to the TensorFlow Channel → bit.ly/TensorFlow1
Get started at → www.tensorflow.org/
Speake: Bill Luan
T2BB62 event: Google I/O 2019; re_ty: Publish; product: Coral - General; fullname: Bill Luan; - Věda a technologie
So in future, in addition to software updates, we should expect "model" update as well ;) This stuff is really cool. Thank you team and Google
It is a well-thought product. I am very impressed with the coral accelerator which can empower the existing edge devices like Rasperry PI
At 27:34 the speaker states: "Be aware that open sourced models are for non-commercial use only." But MobileNetV2 is under an Apache 2.0 License which would not restrict such use. What am I missing here?
Their particular trained model (the weights) is for non commercial use. You can train your own Mobilenet and use that commercially 👍
I was curious how so much can be developed and supported for free but I guess it's the non-profit arm of the mothership.
Very cool, but the Coral USB Accelerator is still quite expensive. It is almost same price as Raspberry Pi4.
Model Play being introduced in this session, is developed by Gravitylink. Model Play is a global platform for discovering, sharing, and experiencing easy to use machine learning models. It’s compatible with Google Coral Dev Board, with a fast, simple, and small control tool in Model Play app that works with AI hardware. Users could start and stop the ML model running, with previewing the performance in real-time, just through by a little mobile phone. Even AI beginners can easily run the model after downloading the Server follow the user guide. It is fast, simple and easy to complete all the setups and to use. learn more at: model.gravitylink.com/
Great stuff guys. The AI space is moving so fast that it is hard to keep up. Maybe an idea, but move those TPU chips to the PC boards.
it's amazing! i've already orderd coral dev board and have a try on model play! that team is genius!
congratulations! how is your experience so far with the coral development board?
@@PALTUBABY we've made an app that could run ML models and see the real-time performance by a camera. super cool !
I downloaded the model play in Google play,it was very easy to run the model very well. Looking forward to more model updates.
I have built a remote controlled lawn mower. Anyone recommend an algorithm to have coral detect objects on the edges of the lawn so the mower stays in the boundary?
Wow! Thank you!
Just ordered the jetson nano and wondering if I should have gotten the coral usb. Can combine both and get a nice dev board plus AI ML on the device. Hoping to find some pulsars.
I got the dual M.2 one.
That’s interesting that they compared this to the Maxwell Nvidia GPUs that are quite old. Still impressive!
ohhhhh～I am very expected
Demo video can be seen at 29:00
I think I will be using this, thank you for that wonderful introduction!!
Is it available online now? Where can I order this?
Thanks
How can I build tensorflow on coral??
Hi do you have an Coral Dev Board install on a Intel NUC 11?
31:35 If you look closely to the display, it identifies all food as “Not Hotdog”, had he placed a hotdog in front of it, it would have blinked Green, followed by yellow and ketchup. However, blinks green and solid red with cracking sound if it’s a Beyond Meat hotdog with a message “Not Hotdog “. Still not GA really.
What kind of applications have most target with this kit? Does Google have a desire to help those tinkering with the Google Coral?
I put the dual m.2 one with a jetson nano 4gb dev kit.
Hello. What do I need to know to build a deep learning library? tell me the courses and books
Deeplearning.ai on CZcams
did you find anything worth? I am interested too
The speed of inference on coral could potentially just be due to the quantization, not their hardware. Its a shame they compared it to a $500 NVIDIA graphics card in an unfair manner, its all marketing. Why don't they run the quantized model on a GPU and see the FPS on that.
It's a piece of trash. It takes a ton of time to set up compared to Jetson nano and raspberry Pi which are almost plug and play.
What's the point of using such Pi-like device in business?