Hi! I was looking into how to dockerize an object detection script I am running on a raspberry pi. Is there a working docker file or documentation that I can follow for this?
We provide a pre-configured Docker image specifically designed for the Hailo AI Suite, which is primarily intended for training models. However, when it comes to running detection and other inference applications, we typically recommend using a virtual environment instead of Docker.
If you still prefer to use Docker for your inference apps, my suggestion would be to use a base image that closely aligns with the current Raspberry Pi OS version you are targeting. This will help ensure optimal compatibility and minimize any potential issues.
As for the requirements and dependencies needed for your Docker setup, you can refer to the hailo-all installation on your Raspberry Pi. It includes all the necessary components and libraries required to run Hailo applications.
Let me know if there’s anything else I can assist you with!
Regards
I see, thank you for the reply. I was looking into the level of effort into containerizing this application as I was a bit concerned as there was a change in the picamera2 library that was not backwards compatible with the previous version of hailo-all. The output of the hailo.run method changed breaking all of my builds and I had to version my code so that any new raspberry pi’s that I install my software on will be able to handle the new output data structure. I tried to manually install the right previous versions of the package directly by using the installers provided from the software suite but could never get the right combination to work on my raspberry pi.
Are there release notes that I can subscribe to so that I am aware of any upcoming breaking changes? And is it possible to version the hailo-all packages so that I can install a previous version if the latest breaks my build?