PyTorch Transfer Learning Image Classifier
This is the final capstone project from the ‘AI Programming with Python’ Nanodegree issued by Udacity. In this final project, I built a software program that preforms Neural Network Transfer Learning using PyTorch.
Project Details
● Project required use of the following: Python, PyTorch, NumPy, Pandas, Matplotlib, Jupyter Notebook, argparse module, command line.
● coded in Python & PyTorch to generate Data Transformations for image data.
● preformed Transfer Learning using PyTorch's torchvision.models (vgg16 & densenet161) to generate & train a Neural Network model on the new Data Transformations.
● model obtained 83% accuracy in 20 epochs. wrote code to save & reload the model’s checkpoint.
● integrated functional programming techniques and utilities to optimize the program.
● built command line application w/Python's argparse module, allowing users to load data & train a new Neural Network model (integrating the functions above). the user can specify: learning rate, hidden units, epochs, GPU usage, save path, & select from PyTorch's torchvision.models.
Link to Project on GitHub