YOLO Inspector

English | 中文

---

An intelligent object detection platform based on Flask and YOLOv8, providing a clean and beautiful user interface with powerful AI recognition capabilities.

Features

• 🎯 Smart Detection: High-precision object detection based on YOLOv8 model
• 🌐 Web Interface: Modern responsive user interface
• 🌍 Multi-language: Automatic language detection (Chinese/English)
• 📁 Drag & Drop: Support drag-and-drop files and click upload
• 📊 Results Display: Real-time display of detection results and confidence statistics
• 🖼️ Visualization: Comparative display of original and annotated result images
• 🔍 Image Zoom: Click to enlarge images for detailed viewing
• 📱 Mobile Friendly: Support for mobile and tablet device access

Interface Preview

YOLO Inspector main interface showcasing modern design and intuitive operation workflow

Interface Features

• Left Operation Panel: Model selection, image upload, language switching
• Right Results Display: Side-by-side comparison of original and detection results
• Smart Statistics Cards: Detected objects count, max confidence, average confidence
• Detailed Detection List: Shows all detected object categories and confidence scores
• Image Zoom Function: Click images to view full-screen high-resolution details
• Responsive Design: Optimized for both desktop and mobile devices

System Requirements

• Python 3.8+
• 8GB+ RAM recommended
• GPU support (optional but recommended)

Installation Guide

1. Clone and enter project directory

cd radar-vision-identify-ai/vision-detect-check

2. Create virtual environment (recommended)

# Create virtual environment
python -m venv venv

# Activate virtual environment
# Windows:
venv\Scripts\activate
# macOS/Linux:
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Verify model file

Ensure the model/best.pt file exists and is accessible. This is the trained YOLOv8 model file.

Quick Start

⚡ If you encounter "ultralytics.utils" error

Fastest solution:

python quick_fix.py

📋 If you encounter model loading issues (PyTorch 2.6+)

Complete fix script:

python fix_torch_compatibility.py

Start server

Method 1: Use smart startup script (recommended)

python run.py

Method 2: Direct startup

python app.py

Access application

Open browser and visit: http://localhost:5800

Usage Instructions

1. Upload Image

• Method 1: Click "Select Image" button to choose file
• Method 2: Directly drag image files to upload area

2. View Results

• System automatically processes uploaded images
• Displays comparison between original and detection result images
• Shows statistics like number of detected objects, confidence levels
• Lists all detected object categories and confidence scores

3. Language Switching

• System automatically detects browser language
• Manual switching via language buttons in top-right corner
• Direct access: /en for English, /zh for Chinese

4. Supported Formats

• Image formats: JPG, PNG, GIF, BMP, TIFF
• File size: Maximum 16MB

API Endpoints

File Upload Recognition

POST /upload
Content-Type: multipart/form-data

Parameters:
- file: Image file

Response:
{
    "success": true,
    "filename": "uploaded_filename",
    "detections": [
        {
            "bbox": [x1, y1, x2, y2],
            "confidence": 0.95,
            "class_id": 0,
            "class_name": "person"
        }
    ],
    "num_detections": 1,
    "original_image": "/uploads/filename.jpg",
    "result_image": "/results/result_filename.jpg"
}

Get Model Information

GET /model_info

Response:
{
    "model_loaded": true,
    "model_path": "model/best.pt",
    "classes": {
        "0": "person",
        "1": "car",
        ...
    },
    "model_type": "YOLO"
}

Language-specific Routes

GET /          # Auto-detect language
GET /en        # Force English
GET /zh        # Force Chinese

Project Structure

vision-detect-check/
├── app.py                    # Flask main application
├── requirements.txt          # Dependencies list
├── README.md                # English project documentation (default)
├── README_ZH.md             # Chinese project documentation
├── run.py                   # Smart startup script
├── model/                   # Model files directory
│   └── best.pt             # YOLOv8 model file
├── templates/              # HTML templates
│   ├── index.html          # Chinese homepage template
│   └── index_en.html       # English homepage template
├── ui/                     # Interface screenshots
│   └── page.png            # Main interface screenshot
├── uploads/                # Upload storage (auto-created)
├── results/                # Result images storage (auto-created)
├── Dockerfile              # Docker configuration
├── docker-compose.yml      # Docker Compose configuration
└── deploy.sh              # Deployment script

Configuration

Model Configuration

• Model path: model/best.pt
• Supported formats: PyTorch (.pt), ONNX (.onnx)

Server Configuration

• Default port: 5800
• Maximum file size: 16MB
• Debug mode: Enabled by default in development

Modify Configuration

You can modify the following configurations in app.py:

# File size limit
app.config['MAX_CONTENT_LENGTH'] = 16 * 1024 * 1024  # 16MB

# Model path
MODEL_PATH = 'model/best.pt'

# Allowed file types
ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif', 'bmp', 'tiff'}

Docker Deployment

Build and run with Docker

# Build image
docker build -t yolo-inspector .

# Run container
docker run -p 5800:5800 yolo-inspector

Use Docker Compose

# Start service
docker-compose up -d

# Stop service  
docker-compose down

Use deployment script

# Make executable
chmod +x deploy.sh

# Deploy
./deploy.sh build
./deploy.sh start

Troubleshooting

Common Issues

1. ultralytics.utils module error

   Problem: See "No module named 'ultralytics.utils'" or similar errors

   Solution:

   Quick fix (recommended):

   python quick_fix.py

   Manual fix:

   pip uninstall ultralytics -y
   pip install --no-cache-dir ultralytics>=8.0.196

2. Model loading failure (PyTorch 2.6+ compatibility issues)

   Problem: If you see errors like "Weights only load failed" or "WeightsUnpickler error", this is a PyTorch 2.6+ compatibility issue.

   Solution:

   Method 1: Use complete fix script (recommended)

   python fix_torch_compatibility.py

   Method 2: Manual PyTorch downgrade

   pip uninstall torch torchvision -y
   pip install "torch>=2.0.0,<2.6.0" "torchvision>=0.15.0,<0.20.0"

   Method 3: Code auto-fix included

   • Latest app.py includes compatibility fix code
   • Run application directly

3. Other model loading issues

   • Check if model/best.pt file exists
   • Confirm ultralytics package is installed correctly
   • Check Python version compatibility

4. Dependency installation failure

   • Upgrade pip: pip install --upgrade pip
   • Use mirror source: pip install -r requirements.txt -i https://pypi.douban.com/simple/

5. GPU related issues

   • CPU mode: Automatically uses CPU for inference
   • GPU mode: Ensure CUDA environment is configured correctly

6. Port occupied

   • Modify port number in app.py: app.run(port=5801)

Performance Optimization

1. GPU Acceleration

   # Install GPU version PyTorch
   pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118

2. Model Optimization

   • Use TensorRT to optimize model
   • Convert to ONNX format to improve inference speed

Development

Extended Features

• Add video file recognition support
• Integrate more pre-trained models
• Add user authentication system
• Implement recognition result database storage

Code Structure

• load_model(): Model loading function
• detect_objects(): Core object detection function
• upload_file(): File upload processing route
• model_info(): Model information query route
• Language detection and routing logic

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

This project is licensed under the MIT License. See LICENSE file for details.

Support

If you encounter issues or have suggestions, please contact us through:

• Create an Issue to report problems
• Submit a Pull Request to contribute code
• Send email for technical questions

---

Note: Please ensure you have the right to use the provided YOLOv8 model files and comply with relevant license requirements.