Google's Coral Dev Board with Edge TPU - realtime object detection at 70fps

I need to to object detection in combination with Augmented reality using ARCore on an Android phone. The reason I will be using ARCore is that object detection alone will not allow you to track objects in 3d space.

With augmented reality I can place markers where objects are detected and therefore count inventory in a warehouse for example. It wasn't practical until now. Streaming to server has too much latency for AR and Android phones are too slow with object detection.

This, tethered to a phone or even better somehow getting the usb TPU dongle to work with an Android phone running a Linux chroot/proot will do the job. Finally. I'm excited to get my hands on one of these.

nice video! I'd use the module to build some basic i/o robots for use around the house :-).

how does it compare to jetson nano? it claims it can handle several 1080p 30fps streams, how well does it performs inference with multiple video streams?

great! Looking forward to seeing more videos to use.

I would help blind people orient themselves in as many environments as possible.

The Google's Coral Dev Board board could use high res stereo camera depth perception like the Stereo Labs ZED camera (it can reach about 40m would be nice to get 100m). Combine object recognition with distance and speech recognition and voice synthesis together and you get tomorrow land.

Pi4 with the TPU gets a smooth 30fps.

Hi. you can share source code?

how much is it, where can you buy it?

I would definitely use it to build a smart intercom, receiving a push notification as well as a picture when a person is at the door.

Is there any way to create some Coral Dev Board rack? Like with the RPs. That would be amazing.

I would use it for an autonomous golf green mower to detect people, animals, golf balls, flags, holes, trees, and even small obstacles lying on a golf green.

Which camera you used to this project and sujjest me better camera for coral dev board.

Put one at the back of a phone, and you won't run into a wall anymore.

Have you know the commands to motorize the TPU in this edge device?

Can we make a phone with it

Thx.
Please tell me how many FPS "coral dev board" will give if the model is only with:
1) person
2) person, cat, dog, bird, car, bicycle
Can you do two tests? Do you have an opportunity?

It seems the smaller USB or M.2 devices are just as fast as the full-size dev board ... what's the advantage or use-case for the full size board?

Can we convert every model to tflite I thought only ssd models and efficientdet like the documentation says??

What is the purpose of the gpios when there is no such library access them for sensor interfacing & many more

What is the video resolution on which you are inferencing in this video?

can we code in python, also what type of algorithms can be used ???

I am an absolute dull/beginner in this field, can I learn?

Hi do you have an Coral Dev Board install on a Intel NUC 11?

I would like to use the board for Number plate detection. Can you confirm if I can run YoloV8 on this board. I got to know that Yolo supports GPU and CPU. So bit confused about the NPU.

What is the resolution?

node red integration

Which one should I choose google coral dev or raspberry pi 5 or jetson nano