← Models

CV / YOLO11

YOLO11-seg

reComputer R Series (CM5 + Hailo-8)

YOLO11 segmentation models (n/s/m/l/x-seg) on reComputer R Series with Hailo-8.

10 downloads
Size
7.9 MB
Memory
4GB+
Precision
Hailo HEF

Choose the device you're using, the set up guide and documentation will update accordingly.

Getting Started

Deploy
sudo docker run --rm \
  --name pi5-hailo8-yolov11n-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11n_seg:latest \
  python web_detection.py --model_path model/yolov11n_seg.hef --video_path video/test.mp4

Model Details

YOLO11-seg on reComputer R Series (CM5 + Hailo-8)

YOLO11n/s/m/l/x-seg run on Hailo-8 through HailoRT 4.23. The HEF exposes ten raw heads; the host decodes DFL boxes, class scores and mask coefficients, assembles sigmoid(coeffs @ proto) instance masks and renders them for the 80 COCO classes.

Model Info

VariantInputPackaged HEFContainer image
YOLO11n-seg640×640×3 RGB7.9 MByolov11n_seg
YOLO11s-seg640×640×3 RGB19.7 MByolov11s_seg
YOLO11m-seg640×640×3 RGB39.6 MByolov11m_seg
YOLO11l-seg640×640×3 RGB40.6 MByolov11l_seg
YOLO11x-seg640×640×3 RGB68.0 MByolov11x_seg

Hardware and Host Setup

ItemValue
BoardreComputer R Series with Raspberry Pi CM5
AcceleratorHailo-8 over PCIe, exposed as /dev/hailo0
RuntimeHailoRT 4.23.x (host / container must share major.minor)
Python in container3.11, aarch64
bash
sudo apt update
sudo apt install hailo-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

Select the command matching the model variant.

YOLO11n-seg

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11n-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11n_seg:latest \
  python web_detection.py --model_path model/yolov11n_seg.hef --video_path video/test.mp4

YOLO11s-seg

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11s-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11s_seg:latest \
  python web_detection.py --model_path model/yolov11s_seg.hef --video_path video/test.mp4

YOLO11m-seg

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11m-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11m_seg:latest \
  python web_detection.py --model_path model/yolov11m_seg.hef --video_path video/test.mp4

YOLO11l-seg

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11l-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11l_seg:latest \
  python web_detection.py --model_path model/yolov11l_seg.hef --video_path video/test.mp4

YOLO11x-seg

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11x-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11x_seg:latest \
  python web_detection.py --model_path model/yolov11x_seg.hef --video_path video/test.mp4

Open http://<Board_IP>:8000 to view the web preview (instance masks overlaid per class plus detection boxes).

USB Camera Mode

bash
sudo docker run --rm \
  --name pi5-hailo8-yolov11n-seg \
  --privileged \
  --net=host \
  -e PYTHONUNBUFFERED=1 \
  --device /dev/hailo0:/dev/hailo0 \
  --device /dev/video0:/dev/video0 \
  -v /usr/lib/libhailort.so.4.23.0:/usr/lib/libhailort.so.4.23.0:ro \
  -v /usr/lib/libhailort.so:/usr/lib/libhailort.so:ro \
  ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11n_seg:latest \
  python web_detection.py --model_path model/yolov11n_seg.hef --camera_id 0

REST API

Each variant has its own endpoint:

VariantEndpoint
YOLO11n-segPOST /api/models/yolov11n_seg/predict
YOLO11s-segPOST /api/models/yolov11s_seg/predict
YOLO11m-segPOST /api/models/yolov11m_seg/predict
YOLO11l-segPOST /api/models/yolov11l_seg/predict
YOLO11x-segPOST /api/models/yolov11x_seg/predict
bash
curl -X POST "http://<Board_IP>:8000/api/models/yolov11n_seg/predict" \
  -F "file=@test.jpg"

Implementation Notes

  • The HEF exposes 10 raw vstreams and no on-chip NMS: per stride (20×20, 40×40, 80×80) a 64-channel DFL box head, an 80-channel class head and a 32-channel mask-coefficient head, plus a 160×160×32 prototype.
  • The host runs the DFL softmax expectation (16 bins), sigmoid on raw class heads, per-class NMS (100 detections in the preview) and sigmoid(coeffs @ proto) mask assembly cropped to each box (threshold 0.5).
  • Letterbox padding is 114; the mean 0 / std 255 normalization is compiled into the HEF. Boxes and masks are mapped back to the frame before drawing.
  • The HEF input size is read at startup (640×640). CPU decode is the preview bottleneck on the larger variants.

Development Notes

  • Sources: src/rpi5_hailo8_yolov11{n,s,m,l,x}_seg/
  • Dockerfiles: docker/hailo8/yolov11{n,s,m,l,x}_seg.dockerfile
  • Containers: ghcr.io/seeed-projects/recomputer-hailo8-cv/yolov11{n,s,m,l,x}_seg:latest
  • HEFs: Hailo Model Zoo v2.20.0 (hailo8), sizes and SHA256 recorded in each module's model/README.md
  • Family: YOLO11-seg (Ultralytics, AGPL-3.0)

Inputs and Outputs

Input: image, video, or USB camera frame. Output: COCO class boxes with instance masks and an annotated MJPEG preview.