Hi everyone,
I have a Raspberry Pi 5 with AI HAT+ (Hailo-8, 26 TOPS) and I’m building a counter-drone system for a college project demo.
I trained a custom YOLOv8n model (1 class: “drone”) and exported it to ONNX, but I’m unable to compile it to .hef because DFC v5.3.0 dropped Hailo-8 support from the CLI, and I don’t have access to the older DFC v3.29.0.
Would someone with a working Hailo-8 compilation setup be willing to compile my ONNX model to .hef?
Model details:
- Architecture: YOLOv8n (nano, ~6MB)
- Classes: 1 (drone)
- Input: 640x640x3
- Format: ONNX, exported with opset=11
- Target: Hailo-8 (not 8L)
Download my ONNX file:
What I need back:
- The compiled .hef file for Hailo-8
Compilation command (should be something like):
hailomz compile yolov8n
–ckpt drone_nano.onnx
–hw-arch hailo8
–classes 1
–performance
The pre-installed yolov8s_h8.hef on my Pi runs at 30 FPS, so I know the hardware works. I just need my custom drone model compiled the same way.
I'd really appreciate the help! Happy to share my project details — it's a pan-tilt drone tracking system with Raspberry Pi 5, ESP32 motor control, thermal camera, and Hailo-8 AI detection.
Thank you!