Dogs vs Cats Classification Project

A deep learning project for classifying dog and cat images using Convolutional Neural Networks (CNN) with TensorFlow and Keras.

Overview

This project implements a binary image classifier to distinguish between dogs and cats, with additional testing on breed classification. The system utilizes a CNN architecture for robust feature extraction and classification.

Project Structure

.
├── dogs_cats.py        # Main implementation class
└── module10.ipynb      # Training and visualization notebook

Features

Dataset Management

• Automated dataset organization into train/validation/test splits
• Built-in data augmentation pipeline
• Integrated TensorFlow dataset creation

Model Architecture

• Custom CNN implementation
• Data augmentation layers
• Optimized for binary classification

Installation

pip install tensorflow matplotlib numpy

Usage

Dataset Preparation

from dogs_cats import DogsCats

dogs_cats = DogsCats()
dogs_cats.make_dataset_folders('validation', 0, 2400)
dogs_cats.make_dataset_folders('train', 2400, 12000)
dogs_cats.make_dataset_folders('test', 12000, 12500)
dogs_cats.make_dataset()

Model Training

dogs_cats.build_network()
dogs_cats.train('model.dogs-cats')

Prediction

dogs_cats.predict('path/to/image.jpg')

Model Architecture

The CNN architecture includes:

• Data augmentation layers
• Convolutional layers with max pooling
• Dense layers for classification
• Binary cross-entropy loss function

Training Process

Callbacks Implementation

• Early stopping for preventing overfitting
• Model checkpointing for saving best weights
• TensorBoard integration for monitoring

Model Persistence

Save Model

dogs_cats.save_model('model.dogs-cats')

Load Model

dogs_cats.load_model('model.dogs-cats')

Performance Visualization

The project includes tools for visualizing:

• Training/validation accuracy
• Loss curves
• Prediction results