🌸 Bangladeshi Flower Classifier

A deep learning-powered web application for classifying 13 different types of Bangladeshi flowers using MobileNetV2 architecture with transfer learning.

🌐 Live Demo

🚀 Try the app now: https://flower-image-classify.streamlit.app/

Upload any flower image and get instant classification results with confidence scores!

🌺 Overview

This project uses a fine-tuned MobileNetV2 model to identify and classify Bangladeshi flowers with an impressive 99.87% validation accuracy. The model was trained on the ColoredFlowersBD dataset containing 13 different flower species commonly found in Bangladesh.

🎯 Supported Flower Classes

1. Chandramallika - Chrysanthemum
2. Cosmos Phul - Cosmos flower
3. Gada - Marigold
4. Golap - Rose
5. Jaba - Hibiscus
6. Kagoj Phul - Bougainvillea
7. Noyontara - Vinca/Periwinkle
8. Radhachura - Flame of the Forest
9. Rangan - Ixora
10. Salvia - Sage flower
11. Sandhyamani - Four o'clock flower
12. Surjomukhi - Sunflower
13. Zinnia - Zinnia flower

🏗️ Model Architecture

• Base Model: MobileNetV2 (pretrained on ImageNet)
• Transfer Learning: Fine-tuned last 50 layers
• Input Size: 224×224×3
• Output: 13 classes (softmax activation)
• Custom Head: GlobalAveragePooling2D → Dropout(0.4) → Dense(13)
• Optimizer: Adam (learning rate: 1e-4)
• Loss Function: Categorical Crossentropy

📊 Model Performance

• Training Accuracy: 99.89%
• Validation Accuracy: 99.87%
• Dataset Split: 80% Training, 20% Validation
• Total Images: 7,927 images
• Training Images: 6,332 images
• Validation Images: 1,595 images

🔧 Data Preprocessing

• Data Augmentation: Rotation, width/height shift, shear, zoom, horizontal flip
• Normalization: Pixel values scaled to [0,1]
• Image Resizing: All images resized to 224×224 pixels
• Batch Size: 32

🚀 Deployment

🔧 Requirements Fix

If you encounter TensorFlow version compatibility issues during deployment, the requirements.txt has been updated with flexible version constraints that work across different Python versions and deployment platforms.

Local Deployment

1. Clone the repository:

   git clone <repository-url>
   cd Image_Classification

2. Install dependencies:

   pip install -r requirements.txt

3. Ensure model files exist:

   • best_model.h5 or flower_classifier.h5
   • class_names.pkl (optional - defaults will be used)

4. Run the Streamlit app:

   streamlit run app.py

5. Open your browser and navigate to http://localhost:8501

Cloud Deployment Options

Streamlit Community Cloud

1. Push code to GitHub repository
2. Connect to Streamlit Community Cloud
3. Deploy directly from GitHub

Heroku

1. Create Procfile:

   web: streamlit run app.py --server.port=$PORT --server.address=0.0.0.0
   

2. Deploy using Heroku CLI or GitHub integration

🛌 Troubleshooting

Common Deployment Issues

TensorFlow Version Compatibility

If you encounter TensorFlow version errors:

1. The requirements.txt uses flexible versioning (tensorflow instead of tensorflow==2.13.0)
2. This allows the deployment platform to choose the compatible version
3. For local development, you may need: pip install tensorflow>=2.15.0

Model File Issues

• Ensure either best_model.h5 or flower_classifier.h5 exists in the project directory
• Model files are large (24MB) - some platforms may have size limits
• Consider using Git LFS for model files in version control

Memory Issues

• TensorFlow models require sufficient RAM (recommend at least 1GB)
• Consider using smaller model architectures for resource-constrained environments

Google Cloud Run

1. Create Dockerfile
2. Build and push container image
3. Deploy to Cloud Run

📁 Project Structure

Image_Classification/
├── Image_Classification_Improved_Model.ipynb  # Main training notebook
├── app.py                                     # Streamlit web application
├── requirements.txt                           # Python dependencies
├── best_model.h5                             # Trained model (24MB)
├── flower_classifier.h5                      # Alternative model file
├── class_names.pkl                           # Class names (optional)
├── dataset/                                  # Dataset folder (excluded from git)
│   ├── train/                               # Training images
│   └── val/                                 # Validation images
└── README.md                                # This file

🖥️ Web Application Features

• 🔍 Real-time Classification: Upload and classify flower images instantly
• 📊 Confidence Scores: View prediction confidence and top-3 results
• 🎨 Beautiful UI: Modern, responsive design with custom CSS
• 📱 Mobile Friendly: Works seamlessly on desktop and mobile devices
• ℹ️ Model Information: Sidebar with model details and supported classes
• 🖼️ Multiple Formats: Supports JPG, JPEG, and PNG image formats

🛠️ Technologies Used

• Deep Learning: TensorFlow/Keras
• Web Framework: Streamlit
• Image Processing: PIL, OpenCV
• Data Manipulation: NumPy, Pandas
• Visualization: Matplotlib
• Model Architecture: MobileNetV2

📈 Training Process

1. Data Collection: ColoredFlowersBD dataset from Kaggle
2. Data Preprocessing: Image augmentation and normalization
3. Model Building: MobileNetV2 with custom classification head
4. Transfer Learning: Fine-tuning approach with frozen initial layers
5. Training: With callbacks (EarlyStopping, ModelCheckpoint, ReduceLROnPlateau)
6. Evaluation: 99.87% validation accuracy achieved

🔄 Model Training Workflow

# Data Augmentation
train_datagen = ImageDataGenerator(
    rescale=1./255,
    rotation_range=30,
    width_shift_range=0.1,
    height_shift_range=0.1,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,
    fill_mode='nearest'
)

# Base Model
base_model = MobileNetV2(weights='imagenet', include_top=False, input_shape=(224, 224, 3))

# Fine-tuning: Unfreeze last 50 layers
for layer in base_model.layers[:-50]:
    layer.trainable = False
for layer in base_model.layers[-50:]:
    layer.trainable = True

# Custom Head
x = base_model.output
x = GlobalAveragePooling2D()(x)
x = Dropout(0.4)(x)
output = Dense(13, activation='softmax')(x)

📊 Performance Metrics

• Loss: Categorical Crossentropy
• Metrics: Accuracy
• Validation Strategy: 20% holdout
• Early Stopping: Patience of 10 epochs
• Learning Rate Reduction: Factor of 0.2 when plateau detected

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📝 License

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

🙏 Acknowledgments

• Dataset: ColoredFlowersBD from Kaggle
• Architecture: MobileNetV2 by Google
• Framework: TensorFlow/Keras team
• Deployment: Streamlit team
• Inspiration: Bangladeshi flora and biodiversity

📞 Contact

For questions, suggestions, or collaboration opportunities, please reach out through the repository issues or discussions.

---

Happy Flower Classification! 🌸🌺🌻