Apple Ripeness Detection

An AI-powered system that automatically detects and classifies apple ripeness levels using YOLOv8. This project helps farmers optimize harvest timing and improve crop management.

Features

• Real-time apple ripeness detection
• Five ripeness levels (20%, 40%, 60%, 80%, 100%)
• User-friendly web interface
• High accuracy (98.9% mAP50)
• Instant processing (~20ms per image)
• Smart recommendations for harvest timing

Installation

1. Clone the repository:

git clone https://github.com/YOUR_USERNAME/AppleRipenessDetection.git
cd AppleRipenessDetection

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows, use: venv\Scripts\activate

3. Install required packages:

pip install -r requirements.txt

Project Structure

AppleRipenessDetection/
├── app.py              # Streamlit web interface
├── predict.py          # Prediction script
├── train.py           # Training script
├── data.yaml          # Dataset configuration
├── datasets/          # Dataset directory
│   └── data/
│       ├── train/    # Training images and labels
│       └── val/      # Validation images and labels
├── predictions/       # Prediction results
├── runs/             # Training results and model weights
└── requirements.txt  # Project dependencies

Running the Project

1. Web Interface (Recommended)

Run the Streamlit app:

streamlit run app.py

Then open your browser and navigate to http://localhost:8501

2. Command Line Prediction

Run predictions on images or directories:

python predict.py

When prompted, enter the path to an image or directory of images.

3. Training the Model

To train the model on your own dataset:

python train.py

Testing the Model

Using Validation Dataset

The project includes a validation dataset with various apple ripeness levels. To test the model:

1. Using Web Interface:

   • Navigate to datasets/data/val/images/
   • Try these sample images:
     • 20%_citra10.jpg (Early growth)
     • 40%_citra13.jpg (Early development)
     • 60%_citra11.jpg (Mid-ripeness)
     • 80%_citra12.jpg (Almost ready)
     • 100%_citra11.jpg (Ready to harvest)

2. Using Command Line:

   # Test single validation image
   python predict.py
   # Enter: datasets/data/val/images/100%_citra11.jpg
   
   # Test all validation images
   python predict.py
   # Enter: datasets/data/val/images/

Expected Results

When testing with validation images, you should see:

• Accurate ripeness level detection
• Confidence scores above 80%
• Clear bounding boxes around apples
• Specific harvest recommendations

Results Location

• Training Results: runs/detect/apple_ripeness_detection9/

  • Best model weights: weights/best.pt
  • Training logs and metrics

• Prediction Results: predictions/

  • Each prediction run creates a new directory
  • Contains processed images and detection information

Model Performance

• mAP50: 98.9%
• mAP50-95: 93.4%
• Precision: 86.7%
• Recall: 93.4%
• Inference Speed: ~20ms per image

Usage Examples

1. Using the Web Interface:

   • Upload an apple image
   • View real-time ripeness analysis
   • Get harvest recommendations

2. Using Command Line:

   # Predict single image
   python predict.py
   # Enter: datasets/data/val/images/100%_citra100.jpg
   
   # Predict directory of images
   python predict.py
   # Enter: datasets/data/val/images/

Requirements

• Python 3.8+
• PyTorch
• Ultralytics YOLOv8
• Streamlit
• OpenCV
• Pillow

License

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