Setting up your new CORAL Edge TPU Dev Board by Google for the first time

[@Moonlight-ip2br]

Can we use a camera that works with the Raspberry Pi or it will not work?

[@ExpatProfessor]

I tried, it did not work for me. The ribbon cable that connects the camera to the board is of different width at the connector and there was no ribbon cable that matches the R Pi to the Coral Dev Board on the market.

[@its_0x4447]

I badly need the coral dev board right now for my ADAS project but the 4GB version is out of stock

[@ExpatProfessor]

Are there none on eBay?

[@kemuri22]

Is this board faster running inference then raspberri pi with a coral tpu accelerator? I'd like to know if there's any sense in switching

[@ExpatProfessor]

From my experience, yes. Depending on resolution, RP does only up to 10 FPS while working along with TPU, compared to some 7 times more (70 FPS) on Coral Edge with TPU.

[@kemuri22]

Raspi4b, 8ram, runs 30 fps on pi camera coral examples. I just wonder if bottle neck is in preview, that comes along with it. Otherwise the tpu is exactly the same in dongle and on the dev board, so it the usb3 connections slowing raspi down compared to dev board?

[@thiruvengadam96]

Can we use Google Coral and feed live footage of a freeway to do vehicle count, ANPR, recognise car make/model?

[@ExpatProfessor]

Yes, once we have a tflite model trained.

[@Moonlight-ip2br]

I downloaded the file in the first and second steps in another computer, and when I tried it in my own computer and wrote mdt devices, my computer did not recognize it. Is there a way to solve it? When I tried Linux, it did not work as well. What is the site and how do I download Linux like the one you have on my HP computer?

[@ExpatProfessor]

Hello, you do need Ubuntu linux, FOCAL 20.04, here is the link to the ISO: releases.ubuntu.com/focal/

[@finnishsnow3182]

Question can the dev board function as its own standalone PC running Linux ? (Using the video output to a standalone monitor).

[@ExpatProfessor]

Yes, it can. This board is already a standalone machine running linux. The linux it is running is similar to ubuntu and is called Mendel Linux. This board is already outputting a video to a standalone monitor. The video that it is outputting is the desktop of Mendel Linux. It is however very plain. All you will see is a desktop picture to make the video beautiful and an icon of the Mendel Linux terminal. That is all. So yes, it is a standalone Linux board with video output already. There is nothing more that you have to do.

[@finnishsnow3182]

@@ExpatProfessor ohhhhh thanks for the reply so I just have to flash the os and then I am good and ready to go right ? I am planning on making a whole system that runs off of solar power. Lmao 🤣 wish me luck 🤞

[@ExpatProfessor]

Yes, that is all that you have to do. Just follow my entire tutorial and you'll end up having linux installed. Wishing you lots of Sun in the sky and good luck!

[@ExpatProfessor]

BTW this board consumes lots of current when it is running an inference model. I bet it is more than 1 A (amp). Perhaps even 3 A.

[@finnishsnow3182]

@@ExpatProfessor HAHAHAHA lmaooooo. Thanks I am looking into alternatives such as Nvidia Jetson nano It is really hard to source any cheap AI dev boards. Do you have any videos regarding the Nvidia Jetson series ?

[@antdx316]

This all doesn't work now? Can you please test it all again?

[@ExpatProfessor]

Coral Edge TPU is dead. Google does not support the hardware any more. It became obsolete.

[@antdx316]

@@ExpatProfessor : ( What did you use yours for now though? Is it a paperweight?

[@antdx316]

@@ExpatProfessor kzbin.info/www/bejne/emqsdquepJ2jrdE Someone got it to work before on v5s.. how can we still do this? We can use Ultralytics HUB to train and then label with it, hopefully it can do API requests when certain things are detected? I did get it to work temporarily but no idea what I did. I cannot figure it out anymore but it was showing bounding boxes.

[@akagordon]

Can the Coral be used for training new machine learning models? I'm interested in using it for lighter machine learning tasks like data mining or maybe grid search to find hyperparameters before passing the model to a GPU.

[@ExpatProfessor]

For training you need a GPU, for instance a laptop with RTX/GTX or you can use in-browser training capabilities provided by google, without your own GPU. For anything else, you can use any machine which runs a decent version of Python.