🏭 RCSIM AI Forge (Hailo Model Conversion Studio)

RCSIM AI Forge is a dedicated toolkit designed for converting and optimizing neural network models from ONNX format to HEF (Hailo Executable Format) binaries. These HEF files are directly executed on Hailo-8 and Hailo-8L AI accelerators (e.g., Raspberry Pi 5 AI Kit) used within the RCSIM autonomous platform.

The project fully automates WSL (Windows Subsystem for Linux) setup, Ubuntu distribution installation, Conda environment configuration (Python 3.10), as well as model compilation and INT8 quantization.

---

📋 System Requirements

• Windows 10/11 Operating System.
• Hailo DFC compiler package placed in the root folder: hailo_dataflow_compiler-3.33.0-py3-none-linux_x86_64.whl.
• Administrator privileges (required only once during WSL installation).

---

🚀 Quick Start

1. Environment Setup:

   • Make sure the Hailo compiler .whl file is in the root directory.
   • Run [1HAILO ENVIRONMENT INSTALLER.bat] (Right-click and select "Run as administrator" if WSL is not already installed on your system).
   • The script will install WSL Ubuntu 22.04, Miniconda, and install the Hailo DFC package.

2. Model Conversion:

   • Copy your .onnx model to the models/ directory.
   • (Optional) Put your calibration data .npy files in the data_tubes/ directory.
   • Run [2Dedicated for Hailo Converter.bat] and follow the numbered interactive command line prompts.

---

📖 User Manual

For a detailed guide on configuration, conversion options (NCHW -> NHWC transposition, INT8 quantization) and troubleshooting, please check the [manual.md] file.