thegarrickchu / springboard-dog_breed_classifier Goto Github PK

Greetings!

This is the repository for my Capstone Project 2 for the Springboard Data Science Career Track. The project/report/deck is entitled "Who’s a good dog?

An exploration and assessment of using Convolutional Neural Networks to classify dog breeds from images".

In this repository, you will find the following folders and files:

1. Main Repo

   1. Final Report

   2. Accompanying Slide Deck

   3. Project 2 Proposal

2. Jupyter Notebooks:

   1. Directory Setup:Verify compressed dataset files, create train test directories and crop images according to annotation. Pickling and packaging images into single Numpy array.

   2. dog_breed_EDA: Exploring a small selection of images, validating data, sizing up the dataset.

   3. First iteration of building CNN from scratch (simple_CNN.ipynb): Initial iterations of CNN models from scratch (before Directory Set up was performed).

   4. CNN from Scratch using Pre-processed image data (simple_CNN_preprocessed_images.ipynb): Final iteration of CNN models from Scratch which became the baseline reference for pre-trained models.

   5. Using Transfer Learning (vgg_model.ipynb): Using VGG16 model architecture and pre-trained weights. Includes visualizations and predictions on test and real-world data.

   6. Deprecated:(simple_CNN_smaller_class_subset.ipynb) Tried using CNN from scratch using flow_from_directory on smaller subset of data.

   7. Deprecated: simple_CNN_v2.ipynb

=======