Q-engineering
Q-engineering
Please note that only TensorFlow 2.4.1 or a lower version will work with the Jetson Nano. Above requires CUDA 11, which is not available for the Jetson Nano.
You cannot run TF 2.13 on a Jetson Nano other than in the CPU mode. If you want GPU, you need to downgrade TF to version 2.4 or lower.
When you first flashed this image, it came with TF 2.4.1 + GPU support installed. Why not start over again, with a new flash?
It recognize the CUDA cores out of the box. There is no need for additional configurations.
Unfortunately, I'm not familiar with the Waveshare carrier board. I'm afraid I can't help you with this issue. One possible workaround may be setting up the system the 'normal' way...
@prpankajsingh, 1) I'm using the ncnn framework running a Yolo deviate, wrapped in a lib. You can use a tailor made TF-lite, sure. 2) Using flaoting points, most networks run...
I'm not sure. Please be so kind and try to install mediapipe to find out. Can you give me the outcome, if you decide to proceed?
From the first glance it looks as if mediapipe will run on the M2-zero. There are no real showstoppers. The only worry is the amount of RAM. 500MB is not...
I understand. If I could, I would like to help you further. However, not the installation of mediapipe is your curl pit, but the deep learning software. With the Armbian...
First, get your app working with a camera. See lines 255,256 in `main.cpp` or the last part of our [site](https://qengineering.eu/deep-learning-examples-on-raspberry-32-64-os.html). Make sure you have `#define AUTO_FILL_DATABASE` active (not comment out)....