a-t-em / binary-image-classification-cat-dog

Convoluted neural network with/without pre-trained base layer to classify cats/dogs.

Cat vs Dog Classifier

This repository contains two models for classifying images of cats and dogs. The model from Kaggle was trained using the Kaggle cat vs dog dataset, which contains ~2,500 images of cats and dogs. The model for the freeCodeCamp project was trained using ~2,000 images provided by freeCodeCamp. Both datasets are balanced.

Kaggle Model
This is a simple CNN model. The model was trained without a pretrained base layer and utilizes a simple helper function to extract image and label data from the Kaggle dataset directory. It achieves an accuracy of around 70% on a held out test dataset of ~5,000 images.

freeCodeCamp Model
This model uses MobileNetV2 as a base layer for a Keras Sequential model. The images were augmented using ImageDataGenerator to improve prediction accuracy. It achieves over 90% accuracy on the 50 images in the testing batch for this project.