I am writing to report that I have successfully launched a dog detection application on a Raspberry Pi 5 running Ubuntu 26.04 and HailoRT 4.24, using ChatGPT.
The following is a report created by ChatGPT, with some modifications made by me.
Hello,
I would like to share a successful user validation result for HailoRT 4.24 on Raspberry Pi 5 with Ubuntu 26.04.
## Test environment
- Raspberry Pi 5
- Ubuntu 26.04 LTS
- Kernel: 7.0.0-1014-raspi
- Hailo-8 AI HAT+
- Architecture: arm64
- HailoRT runtime: 4.24.0
- Hailo PCIe driver: 4.24.0
- Firmware: 4.24.0
- Python: 3.13
- pyHailoRT wheel: 4.24.0
- Camera: Raspberry Pi Camera Module 3
- Model: YOLOv8n HEF
## Installation result
The following components were successfully upgraded from 4.23.0 to 4.24.0:
- hailort
- hailort-pcie-driver
- hailo_pci DKMS module
- Hailo-8 firmware
- Python HailoRT package
After reboot, I confirmed:
```text
hailort 4.24.0
hailort-pcie-driver 4.24.0
hailo_pci 4.24.0
Firmware Version 4.24.0
Kernel 7.0.0-1014-raspi
Device Architecture HAILO8
/dev/hailo0 available
hailort.service active
The following command completed successfully:
hailortcli fw-control identify
Executing on device: 0001:01:00.0
Identifying board
Control Protocol Version: 2
Firmware Version: 4.24.0 (release,app,extended context switch buffer)
Logger Version: 0
Device Architecture: HAILO8
Inference test
I also tested a YOLOv8n HEF with:
hailortcli run ~/Downloads/yolov8n.hef
Running streaming inference (/home/markt/Downloads/yolov8n.hef):
Transform data: true
Type: auto
Quantized: true
Network yolov8n/yolov8n: 100% | 1984 | FPS: 396.31 | ETA: 00:00:00
> Inference result:
Network group: yolov8n
Frames count: 1984
FPS: 396.32
Send Rate: 3895.95 Mbit/s
Recv Rate: 3871.60 Mbit/s
This value is the HailoRT streaming inference result only and does not include camera capture, preprocessing, postprocessing, or JPEG output.
Python dog-detection application
I tested my Python dog-detection application with Camera Module 3.
The application includes:
-
Camera capture
-
Image preprocessing
-
Hailo-8 inference
-
YOLO postprocessing
-
Dog detection
-
JPEG output
Example result:
frame=250 dogs=1 FPS=34.14
frame=251 dogs=1 FPS=32.52
frame=252 dogs=1 FPS=30.20
frame=253 dogs=1 FPS=36.75
frame=254 dogs=1 FPS=35.49
The application operated at approximately 30 to 37 FPS, and dog detection worked correctly.
Kernel warning observation
With HailoRT 4.23.0, I had previously observed repeated find_vma-related kernel warnings when starting the application.
During the initial HailoRT 4.24.0 test period, I did not observe those warnings.
This does not prove that the issue is fully resolved, but the warning was not reproduced in my current test.
Summary
On my system, the following combination is currently working:
Raspberry Pi 5
Ubuntu 26.04 LTS
Kernel 7.0.0-1014-raspi
Hailo-8
HailoRT 4.24.0
Python 3.13
YOLOv8n
Camera Module 3
Dog detection was successful at approximately 30 to 37 FPS.
Please note that this is a user validation result and not a statement of official Ubuntu 26.04 support.
I hope this information is useful to other users testing HailoRT on Ubuntu 26.04.