Skip to content

Flower Classification

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

Flower Classification

README.md

Flower Classification with CNN 🌸

A deep learning project for classifying 102 different flower species using Convolutional Neural Networks (CNN) built with PyTorch.

📋 Table of Contents

🌟 Overview

This project implements a Convolutional Neural Network to classify flower images into 102 different categories. The model is trained on the Oxford 102 Category Flower Dataset and achieves competitive accuracy through careful architecture design and optimization.

Key Features

  • Multi-class classification: 102 flower species
  • CNN Architecture: Custom 3-layer convolutional network
  • Data augmentation: Resize, normalization, and tensor conversion
  • GPU Support: CUDA-enabled training for faster computation
  • Comprehensive evaluation: Per-class and overall accuracy metrics

📊 Dataset

The project uses the Oxford 102 Category Flower Dataset, which contains:

  • 102 flower categories commonly occurring in the United Kingdom
  • 8,189 images total
  • Training/Validation/Test splits pre-defined
  • High-resolution images with varying orientations and scales

Dataset Structure

102flowers/
├── jpg/                    # Image files (image_00001.jpg to image_08189.jpg)
├── imagelabels.mat        # Labels for each image (1-102)
└── setid.mat             # Train/validation/test split indices

Getting the Dataset

The dataset files should be placed in the project directory:

  1. Download the Oxford 102 flowers dataset
  2. Extract images to ./102flowers/jpg/
  3. Place imagelabels.mat and setid.mat in the root directory

🏗️ Model Architecture

CNN Structure

Input (3 x 224 x 224)
    ↓
Conv2d(3→32) + ReLU + MaxPool2d(2x2)     # 32 x 112 x 112
    ↓
Conv2d(32→64) + ReLU + MaxPool2d(2x2)    # 64 x 56 x 56
    ↓
Conv2d(64→128) + ReLU + MaxPool2d(2x2)   # 128 x 28 x 28
    ↓
Flatten → Linear(100352→512) + ReLU       # 512
    ↓
Linear(512→102)                           # 102 classes

Model Specifications

  • Total Parameters: ~51.5M
  • Input Size: 224×224×3 (RGB images)
  • Output: 102 classes (flower categories)
  • Activation: ReLU
  • Loss Function: CrossEntropyLoss
  • Optimizer: SGD (lr=0.001, momentum=0.9)

🚀 Setup & Installation

Prerequisites

  • Python 3.8+
  • CUDA-compatible GPU (optional, but recommended)
  • Git

Installation Steps

  1. Clone the repository

    git clone https://github.com/doroteaMonaco/Flower-Predictor.git
cd Flower-Predictor

  2. Install dependencies

    pip install torch torchvision torchaudio
pip install scipy matplotlib pillow numpy jupyter

  3. For CUDA support (if you have a compatible GPU):

    # Check PyTorch installation with CUDA
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"

  4. Setup dataset (see Dataset section)

💻 Usage

Jupyter Notebook

The main implementation is in Flowers.ipynb. Open it with:

jupyter notebook Flowers.ipynb

Key Code Sections

  1. Data Loading & Preprocessing

    • Load MATLAB files with labels and splits
    • Create custom Dataset class
    • Apply transformations (resize, normalize)

  2. Model Definition

    • 3-layer CNN architecture
    • Forward pass implementation

  3. Training Loop

    • 20 epochs with loss monitoring
    • GPU acceleration (if available)
    • Model checkpointing

  4. Evaluation

    • Per-class accuracy calculation
    • Overall model performance
    • Confusion matrix analysis

Running Training

# In the notebook or Python script
epochs = 20
model = net.cuda() if torch.cuda.is_available() else net

for epoch in range(epochs):
    # Training loop implementation
    # See Flowers.ipynb for complete code

📈 Training

Training Configuration

  • Epochs: 20
  • Batch Size: 16
  • Learning Rate: 0.001
  • Momentum: 0.9
  • Device: CUDA (if available) or CPU

Data Preprocessing

transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

Training Tips

  • GPU Recommended: Training on CPU will be significantly slower
  • Memory Requirements: ~2GB GPU memory for batch size 16
  • Training Time: ~30-60 minutes on modern GPU, several hours on CPU

📊 Results

Performance Metrics

  • Overall Accuracy: [To be updated after training]
  • Training Loss: Monitored per epoch
  • Per-class Accuracy: Detailed breakdown for all 102 flower categories

Model Evaluation

The model provides:

  • Total accuracy across all test samples
  • Individual class accuracies for detailed analysis
  • Confusion matrix for error analysis

📁 File Structure

Flower-Predictor/
├── README.md                 # This file
├── Flowers.ipynb            # Main Jupyter notebook
├── .gitignore              # Git ignore file
├── 102flowers/             # Dataset directory
│   └── jpg/               # Flower images
├── imagelabels.mat        # Flower category labels
├── setid.mat              # Train/test/validation splits
└── model.pth              # Saved model weights (after training)

📦 Requirements

Python Packages

torch >= 1.9.0
torchvision >= 0.10.0
scipy >= 1.7.0
matplotlib >= 3.3.0
Pillow >= 8.0.0
numpy >= 1.21.0
jupyter >= 1.0.0

System Requirements

  • RAM: 8GB+ recommended
  • Storage: 2GB for dataset and dependencies
  • GPU: CUDA-compatible (optional but recommended)

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request. Areas for improvement:

  • Data augmentation techniques
  • Model architecture optimizations
  • Transfer learning implementations
  • Hyperparameter tuning
  • Additional evaluation metrics

Development Setup

  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 the LICENSE file for details.

🙏 Acknowledgments

  • Oxford Visual Geometry Group for the 102 Category Flower Dataset
  • PyTorch Team for the deep learning framework
  • Community contributors and researchers in computer vision

📞 Contact

Author: Dorotea Monaco
GitHub: @doroteaMonaco
Project Link: https://github.com/doroteaMonaco/Flower-Predictor

Built with ❤️ and PyTorch