CV / YOLOX

YOLOX

reComputer R Series (CM5 + Hailo-8)

YOLOX COCO object detection on reComputer RK3576, RK3588, and CM5 with Hailo-8 or Hailo-10H.

246 downloads

Size

5.5 MB

Memory

4GB+

Precision

RKNN / Hailo HEF

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

Platform

Size

Inference Engine

Getting Started

Deploy

sudo docker run --rm \
  --name cm5-hailo8-yolox-tiny \
  --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/yolox_tiny:latest \
  python web_detection.py --model_path model/yolox_tiny.hef --video_path video/test.mp4

REST API

Use the REST API to run inference. Copy the commands below.

Curl

curl http://localhost:8080/v1/chat/completions -d '{
  "model": "yolox-rknn",
  "messages": [{"role": "user", "content": "Hello"}]
}'

Python

import requests

resp = requests.post(
    "http://localhost:8080/v1/chat/completions",
    json={"model": "yolox-rknn", "messages": [{"role": "user", "content": "Hello"}]},
)
print(resp.json())

Model Details

reComputer R Series (CM5 + Hailo-8)

YOLOX on reComputer R Series (CM5 + Hailo-8)

The Hailo-8 YOLOX services provide COCO 80-class object detection with on-chip Hailo HPP NMS. The HEFs emit decoded post-NMS detections, while the FastAPI application provides image prediction, video and camera input, and an MJPEG preview.

Model information

Variant	Input	Packaged HEF	Container image
YOLOX-Tiny	416×416×3 RGB	5.5 MB	`yolox_tiny`
YOLOX-S-Leaky	640×640×3 RGB	9.0 MB	`yolox_s_leaky`
YOLOX-L-Leaky	640×640×3 RGB	53 MB	`yolox_l_leaky`

All three variants use Hailo Model Zoo v2.19.0 Hailo-8 HEFs and have been validated on CM5 with Hailo-8.

Hardware and host setup

Item	Value
Board	reComputer R Series with Raspberry Pi CM5
Accelerator	Hailo-8 over PCIe, exposed as `/dev/hailo0`
Runtime	HailoRT 4.23.x
Python in container	3.11, aarch64

The host driver, firmware, native HailoRT library, and Python wheel in the container must use the same HailoRT major and minor version.

bash

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

hailortcli fw-control identify
ls /dev/hailo0

Run with the demo video

YOLOX-Tiny

bash

sudo docker run --rm --privileged --net=host \
  --name cm5-hailo8-yolox-tiny \
  -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/yolox_tiny:latest \
  python web_detection.py --model_path model/yolox_tiny.hef --video_path video/test.mp4

YOLOX-S-Leaky

bash

sudo docker run --rm --privileged --net=host \
  --name cm5-hailo8-yolox-s-leaky \
  -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/yolox_s_leaky:latest \
  python web_detection.py --model_path model/yolox_s_leaky.hef --video_path video/test.mp4

YOLOX-L-Leaky

bash

sudo docker run --rm --privileged --net=host \
  --name cm5-hailo8-yolox-l-leaky \
  -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/yolox_l_leaky:latest \
  python web_detection.py --model_path model/yolox_l_leaky.hef --video_path video/test.mp4

Open http://<BOARD_IP>:8000 for the Web preview.

REST API

Each variant exposes a model-specific prediction route:

Variant	Endpoint
YOLOX-Tiny	`POST /api/models/yolox_tiny/predict`
YOLOX-S-Leaky	`POST /api/models/yolox_s_leaky/predict`
YOLOX-L-Leaky	`POST /api/models/yolox_l_leaky/predict`

bash

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

The response contains COCO class IDs, confidence scores, and bounding boxes.

Development

Variant	Source	Dockerfile
YOLOX-Tiny	`src/rpi5_hailo8_yolox_tiny/`	`docker/hailo8/yolox_tiny.dockerfile`
YOLOX-S-Leaky	`src/rpi5_hailo8_yolox_s_leaky/`	`docker/hailo8/yolox_s_leaky.dockerfile`
YOLOX-L-Leaky	`src/rpi5_hailo8_yolox_l_leaky/`	`docker/hailo8/yolox_l_leaky.dockerfile`

Inputs and Outputs

Input: image, video frame, or camera frame. Output: COCO classes, confidence scores, and boxes.