This is the completed project for identifying and classifying dog breeds using a Convolution Neural Network (CNN). Submitted as a capstone project for the Udacity Machine Learning Nanodegree.
The main code pipeline is located in dog_app.ipynb and the project report is in final_report.pdf.
A few extra testing images are located in test_images.
The project uses Python 3.5 with the following libraries:
- OpenCV
- Matplotlib
- PyTorch
- Torchvision
The main datasets for this project can be downloaded from here:
- Dogs: https://s3-us-west-1.amazonaws.com/udacity-aind/dog-project/dogImages.zip
- Humans: http://vis-www.cs.umass.edu/lfw/lfw.tgz
The trained models are located here and should be downloaded and saved to the project's root directory:
- Model (Transfer): https://drive.google.com/open?id=1Hc6ojb8iZGsayv1qXuWjdZL4-yArONo9
- Model (Scratch): https://drive.google.com/open?id=1aSWxiUqlz9zZ7IAtFAMy8Q8IGptr1vGu