For working with GPU

[grdsvt]

Hi, I would use with GPU but I have problem with Jetson Nano.
Is it necessary setting something to work with GPU and tensorflow?
I am using the trained model.

Thansk
Regards

[PureHing]

I am not familiar with Jetson Nano and just have used Raspberry Pi. you may need to verify your gpu environment with Jetson Nano Board first, install tensorflow-gpu..., if you can run tensorflow on Jetson Nano well, I think it should work on it.

[grdsvt]

So, you used Raspberry? I suppose Raspberry 4.0
I am using python 3.6.9 with tensorflow 2.1.0. Correct?

[PureHing]

Yes. Raspberry pi4(cpu) with tensorflow object detection API.

[grdsvt]

Hi, I am testing your code with a jetson nano but I have very low fps.......0,5fps.
What is your average fps on Raspberry v.4?

Thnaks.

[grdsvt]

Hi,
I have no good results on jetson nano, so i would try with raspberry.
Please, can you help me?
Which is your fps on raspberry v.4?
Did you change something?

Thanks

[PureHing]

You should optimize the model with TensorRT and run inferencing on the Jetson Nano DevKit. I did not run this repo model on the Raspberry Pi 4, but tested mobilenetv2-ssdlite, which can reach 3 fps, and the CPU usage is 50%.This repo model is smaller than this, and I guess there is a better frame rate.But I think jeston nano with tensorrt is better.If you want to improve the frame rate,Considering model quantification.

[grdsvt]

Hi,
can we evaluate this work together.
I am searching partner to optize the code, can you help me?
Can you send me mail?
Thanks

[PureHing]

Sorry, I don’t have much spare time because of the job is heavily.Just try my best.
MyEmail: xautzxc@gmail.com
Because you already have jeston nano, there is no need to use Raspberry Pi in my opinion, So I said ‘I think jeston nano with tensorrt is better’.
To be honest, I prefer to quantize the model and then choose the Cortex-A series or RISC-V chip to implement my application.