Run DAMO-YOLO on Hailo-10H | reComputer AI Lab

DAMO-YOLO on reComputer CM5 + Hailo-10H

DAMO-YOLO with TinyNAS backbones performs COCO 80-class object detection on Hailo-10H through HailoRT 5.1.1, using the nanodet_split head with DFL box regression. One card, three variants selected via the Size dropdown.

This page targets reComputer R Series (CM5 + Hailo-10H) with a PCIe Hailo-10H accelerator.

Model Info

Variant	Parameters	Operations	mAP (COCO)	Hailo-10H HEF
T (tinynasL20)	11.35M	18.0G	42.8	11.9 MB
S (tinynasL25)	16.25M	37.6G	46.5	21.4 MB
M (tinynasL35)	33.98M	61.6G	49.7	23.9 MB

Property	Value
Task	Object detection
Input	640x640x3 RGB (normalization is a no-op; raw uint8)
Output	6 heads: box (80x80x68, 40x40x68, 20x20x68) + cls (80x80x81, 40x40x81, 20x20x81)
Classes	80 (COCO)
Postprocess	nanodet_split (DFL, regression_length=16), CPU decode
HEF	Hailo Model Zoo v5.4.0, Hailo-10H

The accuracy values are Hailo's Model Zoo reference. They are not benchmarks measured on CM5.

Hardware and Host Setup

Item	Value
Board	reComputer R Series with Raspberry Pi CM5
Accelerator	Hailo-10H over PCIe, exposed as `/dev/hailo0`
Host driver	`hailo-h10-all` apt package
Runtime	HailoRT 5.1.1
Python in container	3.13, aarch64

bash

sudo apt update
sudo apt install hailo-h10-all -y
sudo reboot

# After reboot
hailortcli fw-control identify
ls /dev/hailo0

# Docker
curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh --mirror Aliyun
sudo systemctl enable docker
sudo systemctl start docker

Run With Demo Video

DAMO-YOLO T (tinynasL20)

bash

sudo docker run --rm \
  --name cm5-hailo10h-damoyolo-l20t \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.5.1.1:/usr/lib/libhailort.so.5.1.1:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo10h-cv/damoyolo_tinynas_l20_t:latest \
  python web_detection.py --model_path model/damoyolo_tinynas_l20_t.hef --video_path video/test.mp4

DAMO-YOLO S (tinynasL25)

bash

sudo docker run --rm \
  --name cm5-hailo10h-damoyolo-l25s \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.5.1.1:/usr/lib/libhailort.so.5.1.1:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo10h-cv/damoyolo_tinynas_l25_s:latest \
  python web_detection.py --model_path model/damoyolo_tinynas_l25_s.hef --video_path video/test.mp4

DAMO-YOLO M (tinynasL35)

bash

sudo docker run --rm \
  --name cm5-hailo10h-damoyolo-l35m \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.5.1.1:/usr/lib/libhailort.so.5.1.1:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo10h-cv/damoyolo_tinynas_l35_m:latest \
  python web_detection.py --model_path model/damoyolo_tinynas_l35_m.hef --video_path video/test.mp4

Open http://<Board_IP>:8000 to view the web preview.

REST API

bash

curl -X POST "http://<Board_IP>:8000/api/models/damoyolo_tinynas_l25_s/predict"   -F "file=@test.jpg"

Endpoint	Method	Purpose
`/api/models/damoyolo_tinynas_<variant>/predict`	POST	Detections (JSON)
`/api/video_feed`	GET	MJPEG preview
`/api/config`	GET / POST	Confidence threshold

Development Notes

Source modules: src/hailo10h_damoyolo_tinynas_l20_t/, ..._l25_s/, ..._l35_m/
Dockerfiles: docker/hailo10h/damoyolo_tinynas_l*.dockerfile
Containers: ghcr.io/seeed-projects/recomputer-hailo10h-cv/damoyolo_tinynas_l{20_t,25_s,35_m}:latest
Executor is the HailoRT 5.1.1 multi-output create_infer_model wrapper (damoyolo exposes 6 heads).
HEFs: Hailo Model Zoo v5.4.0 hailo10h builds.

DAMO-YOLO

Getting Started

Model Details

DAMO-YOLO on reComputer CM5 + Hailo-10H

Model Info

Hardware and Host Setup

Run With Demo Video

DAMO-YOLO T (tinynasL20)

DAMO-YOLO S (tinynasL25)

DAMO-YOLO M (tinynasL35)

REST API

Development Notes

Inputs and Outputs