Making a distributable installer

General Discussion
1

Hi all,

I’m building a Tauri application that runs on a Raspberry Pi with a Hailo 8 hat. What’s the best way to setup the Pis so that I can run my own python scripts for object detection?

Currently the rpi5 examples require you to run the .install.sh and then activate an environment. Ideally I’d like to not use an environment and just pre-setup the pi - unless there is a better way to launch an install process without a venv?

Is there a way that I can run the installer and then not have to use the `setup_env.sh`?

Thanks in advance

2

You don’t need to use a virtual environment if you don’t want to. The venv in the examples is just for convenience and isolation. You can run the installer normally and install the required Python packages system-wide, and your scripts should work without sourcing setup_env.sh. Just make sure the HailoRT libraries and Python packages are available in your system environment.

3

Of course I can :man_facepalming:

I see HailoRT and Tappas are available as system packages, but:

  1. Why aren’t the Python bindings (hailort, hailo-tappas-core-python-binding) published on PyPI?

  2. Is hailo-apps-infra actually required for running inference, or is it just for the examples? Can I write detection code without it?

  3. What does hailo-post-install actually configure? Can I pre-configure this in an image?

4

Hi off course you can run your own code w/o the hailo-apps repo.
This repo should be a helpful tool and supply good implementation infrastructure allowing you to concentrate on you code.
Note that it can be pip installed into your project env.
If you are aiming to a high performance pipeline which should run in realtime I suggest you to reuse our pipelines and just connect to output using a callback to your code.