cat-dog-classification

Project Overview

This project documents the end-to-end process of building, training, and evaluating a Convolutional Neural Network (CNN) to classify images of cats and dogs. The project starts with a custom CNN model trained from scratch and includes extensive testing on multiple datasets to evaluate its real-world performance.

The primary goal is to establish a strong baseline model and then use its performance metrics to justify and guide future improvements, such as implementing Transfer Learning.

Key Features

CNN from Scratch: A custom CNN model built with TensorFlow/Keras.

Robust Data Pipeline: Handles large, messy datasets by cleaning and filtering invalid images on the fly.

Checkpointing: Implements a dual-save strategy to save both the best-performing model and the last completed epoch.

Comprehensive Evaluation: The model is evaluated against three different data sources:

A validation set from the original training data.

A large, unseen dataset from Kaggle.

Live, random images fetched from the Unsplash API.

Detailed Reporting: Generates and saves detailed performance reports, including accuracy, precision, recall, and a confusion matrix.