Skip to content

Latest commit

 

History

History

dog-breeds

Dog Breeds Project

This project explores using a convolutional neural network (CNN) to classify different dog breeds using dog images. During prediction if a human image is passed to the model it will attempt to predict what type of dog the person looks like. Transfer learning is used to start with a pre-trained model, specifically the ResNet-50 image classification model which has been trained on ImageNet. The VGG-16 image classification model is also used for transfer learning as an intermediate step.

The data used in this project can be found below:

This project is using Python 3 and needs the following libraries:

  • keras
  • TensorFlow
  • tqdm
  • opencv-python
  • numpy
  • scikit-learn
  • matplotlib
  • jupyter notebook

These can be installed using pip or conda if using Anaconda.

The project is implemented in a Jupyter notebook and can be run using the following from a terminal:

jupyter notebook dog_app.ipynb