Installation of Tensorflow python packages on aml-a311-cc

Greetings everyone,

I am currently using an Alta AML-a311d-cc and connected a USB-camera to it
The OS is Raspbian Bookworm 12 installed from the arm64+arm64.img

So far, I was able to use the USB-camera as a video camera with the python package CV and other methods.

I am currently trying to use a setup to test some video footage with object detection. For this, I am using a python code, which requires tensorflow libraries.

I tried to install it with the command “sudo apt install python3-tensorflow”, but there is no such package available. While searching for a package containing tensorflow like ten, there is no such package.

When trying the command “sudo pip install tensorflow”, I receive the message “This environment is externally managed. To install python packages system-wide, try apt install python3-xyz, where xyz is the package you are trying to install.”

I could bypass this by adding “sudo pip install tensorflow --break-system-package”. However, when trying to install tensorflow this way, I receive the message “Error: Failed building wheel for h5py”. This happens despite “wheel” and “h5py” beeing installed on the device and several reboots after the installation.

Before I try to install some files that might corrupt the current setup I have the following question:
Is there a way to install python packages/libraries like tensorflow on aml-a311-cc that are not currently available in the package distribution?
As the raspbian OS version is a managed OS environment, I am not sure if there are any settings that might interefere with the system with such an installation.

Thank you very much :slight_smile:

Have you read this guide? I think you need to find a way with tflite-runtime.

Only TensorFlow Lite is supported. For that, see @charlie suggested link. TF is not supported. Debian does not package TF or TFLite.

1 Like