This Command Line Image Classifier project was part of my Nanodegree Program with Udacity. This is the first time I've ever created an AI that trains, validates and predicts from scratch.

Mariya Shavshyshvili ([email protected]), with the help and instructions of the Udacity Nanodegree Program.

IMPORTANT CREDITS:

Johnny Metz's video on argparse helped me out a lot with understanding it's uses:

https://youtu.be/cdblJqEUDNo

LearnCode.academy's video on GitHub helped me with uploading this project to GitHub:

https://youtu.be/0fKg7e37bQE

cat_to_name.json

functions_and_classes.py

LICENSE

predict.py

README.md

train.py

Based on project instructions from Udacity, there are a few specific details about the terminal use that are best to note;

in train.py

In the terminal, train.py is best called with:

cd python train.py ./flowers -g

with cd being the directory in which train.py is in.

with -g representing the GPU connection so that everything runs faster.

with ./flowers being the directory in which we save our training, validation and testing sets. The ./flowers directory would have 3 folders: train, valid, test. These folders would be separated into sub-folders, holding the class number of the flower photos they contain.

in predict.py

In the terminal, predict.py is best called with:

cd python predict.py ./flowers/test/11/image_03098.jpg ./save_directory/checkpoint.pth -g

with cd being the directory in which predict.py is in.

with ./flowers/test/11/image_03098.jpg being the path to the image we'd like to predict.

with ./save_directory/checkpoint.pth being the path to the checkpoint we saved in train.py.

with -g representing the GPU connection so that everything runs faster.

As this project was submitted and grated, to avoid direct duplication of the code by future students I won't include the name of the Nanodegree program or the actual name of the project, as stated in the syllabus.