Seeking Guidance on Postprocess Development Environment for Hand Landmark Model with Hailo-8L

Hello,
I’m currently working on a project using the Hailo-8L together with a Raspberry Pi 5, aiming to accelerate inference for a Hand Landmark detection model in real time.

I am using the .hef file provided by the Hailo Model Zoo, but since this particular model is not officially supported, I intend to write a custom postprocess.so file for integration into the TAPPAS pipeline.

Current Status:

  • Using .hef from Model Zoo
  • Planning to build a custom postprocess C++ file and compile it into a shared object (.so)
  • Considering whether to set up TAPPAS or the full Hailo AI Software Suite

Questions:

  1. What is the minimum required development environment for writing and compiling a custom postprocess for Hailo-8L?
  • Is TAPPAS alone sufficient, or is it necessary to install the entire Hailo AI SW Suite?
  1. Are there any official guidelines, documentation, or community examples available for implementing a custom postprocess (e.g., C++ examples, meson.build, ninja usage, etc.)?

Any example or reference on how to implement and integrate a custom postprocess would be greatly appreciated.

Thank you in advance!

Hi @20246145
Welcome to the Hailo community. At DeGirum. we developed PySDK, a python package to simplify application development with Hailo. We integrated the postprocessing for a variety of models so that end users can easily develop using these models. One such model we integrated is the Hand landmark detection model. You can use the model along with our PySDK. The model is available on our AI Hub: DeGirum AI Hub. Once you register for our hub, you can download the model assets. You can see PySDK usage examples at : DeGirum/hailo_examples: DeGirum PySDK with Hailo AI Accelerators

1 Like

Thank you very much for the quick and informative response!
The DeGirum PySDK and pre-integrated hand landmark model sound like a great solution for simplifying development.

I have a follow-up question:
Is it possible to run the PySDK-based application on a Raspberry Pi 5 together with the Hailo-8L?
If so, could you please share any example project, installation guide, or documentation specifically targeting this environment (e.g., setup steps, dependencies, performance considerations)?

Looking forward to trying this out on my setup.
Thanks again for your support!

Hi @20246145
Yes, PySDK works with Raspberry Pi + Hailo8L. The link I shared ( DeGirum/hailo_examples: DeGirum PySDK with Hailo AI Accelerators) contains detailed instructions on setup as well as various examples. Let us know if you run into any issues.