YOLOv26 Dual USB Camera Image Processing System on Jetson

This wiki demonstrates how to build a dual USB camera image processing system based on YOLOv26 model and TensorRT acceleration from scratch. The system includes the following features:

Dual Camera Parallel Processing: Real-time video stream processing for two USB cameras simultaneously
Multi-task Vision Analysis: Object detection, pose estimation, image segmentation (SAM model)
TensorRT Acceleration: Significantly improve inference speed using NVIDIA TensorRT engine
Web Interface Preview: View processing results in real-time through a browser
MJPEG Compression: Reduce USB bandwidth usage and improve system stability

Prerequisites

reComputer J4012 or Other products of the reCompuer series
2 USB cameras
USB Type-C data transmission cable(Connect the camera)

Download from GitHub

Method 1: Clone the Entire Repository (Recommended)

This is the simplest method to get all files at once:

# 1. Navigate to your home directory
# ！！Replace this with your own username
cd /home/seeed

# 2. Clone the repository
git clone https://github.com/bleaaach/yolov26_jetson.git

# 3. Navigate to the project directory
cd yolov26_jetson

# 4. View downloaded files
ls -la

You should see the following files:

run_dual_camera_docker.sh - Docker deployment script
run_dual_camera_local.sh - Local deployment script
README.md - This documentation file

Method 2: Download Individual Files

If you only need specific script files, you can download them individually using wget:

# 1. Navigate to your home directory
cd /home/seeed

# 2. Create project directory
mkdir -p yolov26_jetson
cd yolov26_jetson

# 3. Download Docker script
wget https://raw.githubusercontent.com/bleaaach/yolov26_jetson/main/run_dual_camera_docker.sh

# 4. Download Local script
wget https://raw.githubusercontent.com/bleaaach/yolov26_jetson/main/run_dual_camera_local.sh

# 5. Download README documentation
wget https://raw.githubusercontent.com/bleaaach/yolov26_jetson/main/README.md

# 6. View downloaded files
ls -la

Verify Downloaded Files

After downloading, verify that the files exist:

# View file list
ls -la

# Check file permissions

If the scripts don't have execute permissions, add them:

# Add execute permissions
chmod +x run_dual_camera_docker.sh
chmod +x run_dual_camera_local.sh

# Check again
ls -la

Comparison of Two Deployment Methods

Feature	Docker Method	Local Method
Environment Isolation	✅ Fully isolated, doesn't pollute host environment	❌ Directly installed on host
Deployment Speed	✅ Fast, one-click startup	❌ Requires manual installation of many dependencies
Hardware Access	⚠️ Requires device mapping configuration	✅ Direct access to all hardware
Performance	⚠️ Has container overhead	✅ Better performance
Storage Space	⚠️ Requires ~2GB Docker image	✅ Less storage usage
Recommended Use Case	Quick testing, multi-device deployment	Production environment, best performance

（Option 1）Local Environment Setup

Step 1. Update Package List and Install pip

# Update package list
sudo apt update

# Install pip
sudo apt install python3-pip -y

# Upgrade pip
pip install -U pip

If pip is not pre-installed on the system, use the following command to install:

# Download get-pip.py
curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py

# Install pip
python3 get-pip.py --user

Step 2. Install Ultralytics Package

Install Ultralytics and its optional dependencies (for model export):

# Install Ultralytics
~/.local/bin/pip install ultralytics[export]

Step 3. Install PyTorch and Torchvision

Important: PyTorch and Torchvision installed via pip are not compatible with Jetson's ARM64 architecture. You need to manually install versions built specifically for Jetson.

First uninstall incompatible versions:

# Uninstall incompatible versions
~/.local/bin/pip uninstall torch torchvision -y

Then install JetPack 6.1 compatible versions:

# Install PyTorch 2.5.0
~/.local/bin/pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.5.0a0+872d972e41.nv24.08-cp310-cp310-linux_aarch64.whl

# Install Torchvision 0.20
~/.local/bin/pip install https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.20.0a0+afc54f7-cp310-cp310-linux_aarch64.whl

If GitHub download is slow, you can use an acceleration proxy:

# Use gh proxy to download PyTorch
~/.local/bin/pip install https://gh-proxy.com/https://github.com/ultralytics/assets/releases/download/v0.0.0/torch-2.5.0a0+872d972e41.nv24.08-cp310-cp310-linux_aarch64.whl

# Use gh proxy to download Torchvision
~/.local/bin/pip install https://gh-proxy.com/https://github.com/ultralytics/assets/releases/download/v0.0.0/torchvision-0.20.0a0+afc54f7-cp310-cp310-linux_aarch64.whl

Step 4. Install cuSPARSELt

cuSPARSELt is a dependency of PyTorch 2.5.0 and needs to be installed separately:

# Install cuSPARSELt
sudo apt-get install -y libcusparselt0

Step 5. Install onnxruntime-gpu

onnxruntime-gpu is used for some model export functions. Since the package on PyPI doesn't contain aarch64 binaries for Jetson, manual installation is required:

# Install onnxruntime-gpu 1.23.0
~/.local/bin/pip install https://gh-proxy.com/https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.23.0-cp310-cp310-linux_aarch64.whl

Or use version 1.20.0:

# Install onnxruntime-gpu 1.20.0
~/.local/bin/pip install https://gh-proxy.com/https://github.com/ultralytics/assets/releases/download/v0.0.0/onnxruntime_gpu-1.20.0-cp310-cp310-linux_aarch64.whl

Step 6. Configure PATH Environment Variable

Since user installation mode is used, pip installed executables are located in the ~/.local/bin/ directory. It's recommended to add this directory to the PATH environment variable:

# Add to .bashrc
echo 'export PATH=$PATH:~/.local/bin' >> ~/.bashrc

# Reload .bashrc
source ~/.bashrc

Step 7. Verify Installation

Check installed package versions:

# Check versions
python3 -c "import ultralytics; import torch; import torchvision; import onnxruntime; print('ultralytics version:', ultralytics.__version__); print('torch version:', torch.__version__); print('torchvision version:', torchvision.__version__); print('onnxruntime version:', onnxruntime.__version__)"

Expected output:

ultralytics version: 8.4.7
torch version: 2.5.0a0+872d972e41.nv24.08
torchvision version: 0.20.0a0+afc54f7
onnxruntime version: 1.23.0

Step 8. Test YOLOv26 Inference Functionality

from ultralytics import YOLO
import torch

# Check if CUDA is available
print(f"CUDA available: {torch.cuda.is_available()}")
if torch.cuda.is_available():
    print(f"CUDA device count: {torch.cuda.device_count()}")
    print(f"CUDA device name: {torch.cuda.get_device_name(0)}")

# Load YOLOv26n model
model = YOLO('yolo26v26n.pt')
print(f"Model loaded successfully!")

# Perform inference test
results = model('https://ultralytics.com/images/bus.jpg')
print(f"Inference successful! Detected {len(results[0].boxes)} objects")

# Display detection results
for i, box in enumerate(results[0].boxes):
    cls_id = int(box.cls[0])
    conf = float(box.conf[0])
    cls_name = model.names[cls_id]
    print(f"  Object {i+1}: {cls_name} (confidence: {conf:.2f})")

Step 9. Prepare Model Files

Ensure model files are downloaded to the correct location:

# Check model directory
ls -la /home/seeed/ultralytics_data/

# Create directory if it doesn't exist
mkdir -p /home/seeed/ultralytics_data

If model files don't exist, download them first:

# Navigate to model directory
cd /home/seeed/ultralytics_data

# Download object detection model
yolo export model=yolov26n.pt format=engine device=0

# Download pose estimation model
yolo export model=yolov26n-pose.pt format=engine device=0

# Download segmentation model
yolo export model=yolov26n-seg.pt format=engine device=0

# Verify model files
ls -la

You should see the following files:

yolo26n.engine
yolo26n-pose.engine
yolo26n-seg.engine

Step 10. Run Local Script

Now you can run the Local script:

# 1. Navigate to project directory
cd /home/seeed/yolov26_jetson

# 2. Ensure script has execute permissions
chmod +x run_dual_camera_local.sh

# 3. Run Local script
./run_dual_camera_local.sh

（Option 2）Docker Environment Setup

Step 1. Ensure Docker is Installed

First check if Docker is installed:

# Check Docker version
docker --version

If Docker is not installed, install it first:

# Update package list
sudo apt update

# Install Docker
sudo apt install docker.io -y

# Start Docker service
sudo systemctl start docker

# Add current user to docker group
sudo usermod -aG docker $USER

# Re-login to apply changes
newgrp docker

For more details on configuring Docker environment, please refer to this GitHub repository: https://github.com/zibochen6/reComputer-Jetson-for-Beginners/tree/main/3-Basic-Tools-and-Getting-Started/3.7-Docker/

Step 2. Ensure Docker Service is Running

# Check Docker service status
sudo systemctl status docker

If Docker is not running, start it:

# Start Docker service
sudo systemctl start docker

Step 3. Prepare Model Files

Ensure model files are downloaded to the correct location:

# Check model directory
ls -la /home/seeed/ultralytics_data/

# Create directory if it doesn't exist
mkdir -p /home/seeed/ultralytics_data

If model files don't exist, download them first:

# Navigate to model directory
cd /home/seeed/ultralytics_data

# Download object detection model
yolo export model=yolov26n.pt format=engine device=0

# Download pose estimation model
yolo export model=yolov26n-pose.pt format=engine device=0

# Download segmentation model
yolo export model=yolov26n-seg.pt format=engine device=0

# Verify model files
ls -la

You should see the following files:

yolo26n.engine
yolo26n-pose.engine
yolo26n-seg.engine

Step 4. Run Docker Script

Now you can run the Docker script:

# 1. Navigate to project directory
cd /home/seeed/yolov26_jetson

# 2. Ensure script has execute permissions
chmod +x run_dual_camera_docker.sh

# 3. Run Docker script
./run_dual_camera_docker.sh

Step 5. Access Web Interface

Open the following address in your browser:

http://localhost:5000

At this moment the browser will display the real-time feeds from both cameras side-by-side, similar to “Left Camera Feed | Right Camera Feed”. The actual frame rate and resolution depend on your hardware performance.

note

If you need to restart the system:

# 1. Stop and remove existing container
docker rm -f dual-camera-system

# 2. Re-run script
cd /home/seeed/yolov26_jetson
./run_dual_camera_docker.sh

Resources

Tech Support & Product Discussion

Thank you for choosing Seeed Studio products! For technical support and product discussion, please use the following channels:

Prerequisites​

Download from GitHub​

Method 1: Clone the Entire Repository (Recommended)​

Method 2: Download Individual Files​

Verify Downloaded Files​

Comparison of Two Deployment Methods​

（Option 1）Local Environment Setup​

（Option 2）Docker Environment Setup​

Resources​

Tech Support & Product Discussion​

Prerequisites

Download from GitHub

Method 1: Clone the Entire Repository (Recommended)

Method 2: Download Individual Files

Verify Downloaded Files

Comparison of Two Deployment Methods

（Option 1）Local Environment Setup

（Option 2）Docker Environment Setup

Resources

Tech Support & Product Discussion