🐱🐶 Cat vs Dog Image Classifier

A deep learning CNN model for binary image classification using TensorFlow and Keras

Features • Installation • Usage • Results • Architecture

---

📋 Table of Contents

• Overview
• Features
• Model Architecture
• Installation
• Dataset Structure
• Usage
• Results
• Technical Details
• Contributing
• License

---

🚀 Overview

This project implements a Convolutional Neural Network (CNN) for classifying images of cats and dogs with high accuracy. The model leverages both Transfer Learning with MobileNetV2 and a custom CNN architecture, featuring advanced data augmentation techniques and robust training strategies.

🎯 Key Highlights

• Dual Architecture Support: Transfer Learning (MobileNetV2) + Custom CNN
• Advanced Data Augmentation for improved generalization
• Smart Training Strategies: Early stopping, learning rate reduction
• Production Ready: Complete preprocessing and prediction pipeline
• Comprehensive Visualization of training metrics

---

✨ Features

Feature	Description
🧠 Transfer Learning	Pre-trained MobileNetV2 for superior performance
🔄 Data Augmentation	Horizontal flip, zoom, width/height shifts
📊 Smart Callbacks	Early stopping and adaptive learning rate
🎨 Visualization	Training history plots and performance metrics
⚡ Fast Inference	Optimized for quick predictions
🛡️ Regularization	Dropout layers prevent overfitting

---

🏗️ Model Architecture

Transfer Learning Architecture (Default)

MobileNetV2 (ImageNet weights, frozen)
           ↓
   GlobalAveragePooling2D
           ↓
      Dropout(0.3)
           ↓
    Dense(128, ReLU)
           ↓
      Dropout(0.5)
           ↓
   Dense(1, Sigmoid)

Custom CNN Architecture (Alternative)

Input(160×160×3)
       ↓
Conv2D(16) + BatchNorm + MaxPool + Dropout(0.25)
       ↓
Conv2D(32) + BatchNorm + MaxPool + Dropout(0.25)
       ↓
Conv2D(64) + BatchNorm + MaxPool + Dropout(0.25)
       ↓
GlobalAveragePooling2D
       ↓
Dense(32, ReLU) + Dropout(0.5)
       ↓
Dense(1, Sigmoid)

---

🔧 Installation

Prerequisites

• Python 3.8+
• CUDA-compatible GPU (recommended)

Quick Setup

# Clone the repository
git clone https://github.com/Wayn-Git/CatvsDog.git
cd CatvsDog

# Install dependencies
pip install tensorflow matplotlib seaborn numpy

Dependencies

tensorflow>=2.8.0
matplotlib>=3.5.0
seaborn>=0.11.0
numpy>=1.21.0

---

📁 Dataset Structure

CatvsDog/
├── Data/
│   ├── train/
│   │   ├── cats/          # Training cat images
│   │   └── dogs/          # Training dog images
│   └── test/
│       ├── cats/          # Test cat images
│       └── dogs/          # Test dog images
├── Model/
│   └── cat_dog_model.keras
├── Notebook/
│   ├── CatvsDog/
│   ├── (Backup).ipynb
│   └── catDog(Col...).ipynb
├── PythonScript/
│   ├── catdog.py          # Main training script
│   ├── catDog.ipynb
│   ├── README.md
│   └── requirements.txt
└── README.md

---

🚀 Usage

Training the Model

# Set transfer learning mode
TRANSFER_LEARNING = True  # Use MobileNetV2 (recommended)
# TRANSFER_LEARNING = False  # Use custom CNN

# Run training
python PythonScript/catdog.py

Making Predictions

from tensorflow.keras.preprocessing import image
from tensorflow.keras.models import load_model
import numpy as np

# Load trained model
model = load_model('Model/cat_dog_model.keras')

# Load and preprocess image
img_path = "your_image.jpg"
img = image.load_img(img_path, target_size=(160, 160))
img_array = image.img_to_array(img) / 255.0
img_array = np.expand_dims(img_array, axis=0)

# Make prediction
prediction = model.predict(img_array)
result = "Dog 🐶" if prediction[0][0] > 0.5 else "Cat 🐱"
print(f"Predicted: {result}")
print(f"Confidence: {prediction[0][0]:.4f}")

---

📈 Results

Model Performance

Architecture	Test Accuracy	Training Time	Model Size
Transfer Learning (MobileNetV2)	~95%+	~15-20 epochs	~9MB
Custom CNN	~85-90%	~25-30 epochs	~3MB

Training Features

• Early Stopping: Prevents overfitting with patience=5
• Learning Rate Reduction: Adaptive LR with factor=0.2, patience=3
• Data Augmentation: Improves generalization significantly
• Batch Processing: Efficient training with batch_size=64

---

🔧 Technical Details

Data Preprocessing

# Training augmentation
trainGen = ImageDataGenerator(
    rescale=1./255,
    horizontal_flip=True,
    zoom_range=0.1,
    width_shift_range=0.1,
    height_shift_range=0.1
)

# Test preprocessing
testGen = ImageDataGenerator(rescale=1./255)

Hyperparameters

Parameter	Value
Input Size	160×160×3
Batch Size	64
Optimizer	Adam
Loss Function	Binary Crossentropy
Max Epochs	30
Early Stopping Patience	5
LR Reduction Factor	0.2

Key Features

• Robust Architecture: Both transfer learning and custom CNN options
• Smart Callbacks: Early stopping and learning rate scheduling
• Data Augmentation: Comprehensive image transformations
• Regularization: Strategic dropout placement
• Efficient Training: Optimized batch processing

---

📊 Visualization

The model automatically generates training visualizations:

• Accuracy Curves: Training vs Validation accuracy over epochs
• Loss Curves: Training vs Validation loss progression
• Performance Metrics: Detailed evaluation results

---

🤝 Contributing

Contributions are welcome! Please feel free to:

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

Areas for Contribution

• Model architecture improvements
• Additional data augmentation techniques
• Web interface development
• Mobile deployment optimization
• Documentation enhancements

---

📄 License

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

---

🙏 Acknowledgments

• TensorFlow/Keras team for the excellent deep learning framework
• MobileNetV2 architecture for efficient transfer learning
• ImageNet dataset for pre-trained weights
• Open source community for inspiration and support

---

⭐ If you found this project helpful, please give it a star! ⭐

Made with ❤️ and Deep Learning

Report Bug • Request Feature • Documentation