Received my Raspi 5 and Hailo HAT couple days ago. I’ve been playing around with the example code from the “hailo-rpi5-examples” repo. But now I want to make my own object detection application.
Question is, what do I use?
When I look this up on the forums I see shashi pushing to users to use Degirum.
When I go and see custom projects made from Hailo’s 2024-2025 hackathon most used the “hailo-apps-infra“ repo.
Others either use hailo_platform and HailoRT.
Is there not a centralized Python SDK or library that everyone agrees to use? Do I have to clone a specific repo and pick bits and pieces from it to build my application?
The Model Build environment on the left is used to convert a model into the HEF format that is required to run it on a Hailo device. It runs on x86 on Ubuntu. I recommend installing the Hailo AI Software suite Docker.
The Runtime environment on the right can run on different hosts (x86 or Arm), different OS (Windows, Linux, Yocto) and you can use many different frameworks (GStreamer, OpenCV, ONNX Runtime …) and languages (C/C++, Python) depending your requirements. The choice is yours.
A single “one-size-fits-all” SDK would probably be too restrictive or endlessly complicated.
Instead, we provide the building blocks so people can work in the language and environment that best fits their needs. If everyone had to agree on one central library… well, it might be simpler, but also a lot less flexible (and honestly, a little boring).
