This repository is a stand-alone one for holding my working code developed in during participation in the Udacity Pytorch Challenge site https://github.com/udacity/pytorch_challenge.
I was a participant in the inaugural challenge of 2018/2019, and was successful in being accepted for the Facebook Scholarship that was awarded to 300 of the 10,000 participants in the challenge.
The notebook "image classifier project" was the original template to develop our project, however I have developed my working solution in the other notebooks contained in this repo, then applied them to the notebook. I saved a backup of the original notebook as "image classifier project original".
The challenge required us to develop a flower classifier using transfer learning techniques with Pytorch.
You would need to clone the repo then open the notebooks in Jupyter Notebook, starting with the original "image classifier project" which contains explanations of the data and the purpose of the notebook.
Then, open the other notebooks to see how I worked up solutions to the questions asked in the main notebook. My order of notebook development was "Image Classifier Project-Working-Version-2" where I first explored the problems, after which came "Image Classifier Project-Working-Resnet152" where I developed the working solution utilizing the Resnet152 ImageNet architecture, then "Resnet152_classification" where I further worked on creating a submission to a kaggle competition that was run concurrently with the Pytorch Challenge. The notebook "Image Classifier Project-Working-InceptionResnetV2" is a failed attempt to get an InceptionResnetv2 transfer learning model working.
These are here primarily for me to maintain my own code, but open for others to check and get ideas from or contribute to, provided that anyone using code directly attributes it either to me or to someone I have attributed to if my code has been borrowed. Anyone participating in a Udacity course and especially the same course must ensure that they do not copy directly as Udacity are very strict about checking for plagiarism. While I don't expect anyone to contribute to these notebooks any feedback is appreciated!