How to install tensorflow or tensorflow-lite on raspberry Pi4?

Question

I have a raspberry Pi4 containning ubuntu 18.04. My ubuntu has been installed from a dockerfile and push thanks to balena OS in my raspberry.

Then I tried to install tensorflow for openvino project on it by following this tuto in an environment with python3.7.5: https://qengineering.eu/install-tensorflow-2.1.0-on-raspberry-pi-4.html

But I get the following error: ERROR: tensorflow-2.1.0-cp37-cp37m-linux_armv7l.whl is not a supported wheel on this platform.

Is it possible that balena OS stop the detection of the raspberryPi4 architecture and then the installation ?

If not do you have a good version of tensorflow for raspberrypi4 with ARMv8 architecture.

I don't understand where this problem come from so if you have any idea, thanks a lot.

Answer 1

If I understand your question correctly, you want to install TensorFlow on Docker running Ubuntu 18.04 on RPI4 with balena OS as host.

You could try to build TensorFlow from source as described in their Build from source instructions. Based on latest OpenVINO™ toolkit release (v2020.3), the TensorFlow version that gets installed is v1.14, just keep that in mind when you decide which TensorFlow version to build from source. Also the build process may take a while.