Integrating Detection code with Hailo M.2 on Raspberry Pi 5

I am using a Raspberry Pi 5 with the Hailo M.2 HAT+ kit. I have a segmentation code for detecting bends, and I want to integrate it with the Hailo M.2 to accelerate my code.How should I do that in my code?

Welcome to the Hailo Community!

That is a broad question. The Hailo-8 is a specialized hardware that requires to follow a workflow to make use of it. I will not be able to explain everything in a forum post. I can provide you some pointers to get you started.

As you can see from the image below, to run a neural network on a Hailo device it must be converted from a TFLite or ONNX file into a binary file called HEF (Hailo Executable File) in a multi step process using the Hailo Dataflow Compiler.

The Hailo Dataflow Compiler comes with build in tutorials.

Before you start with this I would recommend you work with the examples for the Raspberry Pi and our prebuilt models to get a good understanding of running a HEF file, sending images to it and post processing the results.

As a next step I recommend retraining an existing model using our retraining dockers and convert the model using the Hailo Model Zoo and run it with the example applications.

Then you can take on the task to convert and integrate your own model.

If you need any help on the way, your best chance to get good advice is to ask very specific questions and provide as much information as possible including your experience level and what you have tried so far.