dog-breed-image-classifier

This repository contains the console example of implementing a pretrained convolutional neural network to check dog breeds by uploaded images. The program was created during Udacity's AI Programming with Python Nanodegree Program.

Dependencies

The next points are necessary for correct execution:

• Python 3.6 or later
• Torch 0.4.0
• Pillow 5.2.0
• torchvision 0.2.1

Please, check your environment.

Quickstart

check_images.py is the main script in this repository. You can run it without any parameters with command:

python check_images.py

And got a result with default command line arguments:

You can uncomment a few lines to print additional information in check_images.py:

• line 29 for command line arguments

• line 43 for checking dog breed's keys

• line 58 for the first step of classification

• line 72 for classified images with label flags

• line 84 for short resume

Command Line arguments

If you want you can manage command line arguments --dir, --arch and --dognames.

Directory with images

The argument --dir defines a directory with images to recognise. By default pet_images folder is used. You can create your own folder, for example uploaded_images, and upload different images there, then run check_images.py with directory path after the optional argument --dir:

python check_images.py --dir uploaded_images/

Neural network architecture

You can choose one architecture from ResNet, AlexNet and VGG to check which one is the most effective for you using --arch. By default ResNet is choosen. This is the example of using VGG:

python check_images.py --dir uploaded_images/ --arch='vgg'

The list of valid dog's breeds

You can change the list of breeds by creating another .txt file and using argument --dogfile. By default the program uses dognames.txt. Please check your file has the same structure as dognames.txt. Usage of another list:

python check_images.py --dir uploaded_images/ --arch='alexnet' --dogfile='breeds.txt'

Issues

• The classifier works with JPEG images
• Images should begin with an object's name and contain only Latin alphabet, digits and underscore symbols. For example Irish_setter_02.jpeg,cat.jpg,german_shepherd_86547.jpg.

License

The contents of this repository are covered under the MIT License.