Developing an Image Classifier with Deep Learning

Introduction:

The goal was to train an image classifier to recognize different species of flowers. Used Udacity’s workspace enabled with GPU for working on this project.

Table of Contents:

This project contains the following files:

• Image Classifier Project.ipynb : This python file has all the code that was used to put through the entire analysis.
• train.py : This includes the python code used to train the deep learning classifier.
• predict.py : This has the code use to predict the images.
• program.py : This contatins all the functions that are used by train.py and predict.py
• output_log.txt: This contains the output log.

In the Terminal or Command Prompt, navigate to the folder containing the project files, and then use the command jupyter notebook Image Classifier Project.ipynb to open up a browser window or tab to work with your notebook. Alternatively, you can use the command jupyter notebook or ipython notebook and navigate to the notebook file in the browser window that opens.

Summary of Analysis:

The process is broken into multiple steps

• Load and preprocess the image dataset Applied transformations such as random scaling, cropping and flipping. Have split dataset into training, validation and test data set.

• Building and training the classifier Loaded a pre-trained network i.e. VGG16 network. Defined a new untrained feed forward network as a classifier, using ReLu activations and dropout. Trained the classifier using the pre-trained network to get the features. Tracked the loss and accuracy on the validation set to determine the best hyper parameters.

• Save and load the checkpoint As the network has been trained, the model was saved so that it can be loaded later for making predictions.

• Class Prediction Used the trained network for interpreting the inference. Pass an image into the network and predict the class of the image.

Software:

This project uses the following software and Python libraries:

• [Python 2.7] (https://www.python.org/download/releases/2.7/)
• [NumPy] (http://www.numpy.org/)
• [Pandas] (http://pandas.pydata.org/)
• [scikit-learn] (http://scikit-learn.org/stable/)
• [matplotlib] (http://matplotlib.org/)

You will also need to have software installed to run and execute a [Jupyter Notebook] (http://ipython.org/notebook.html) If you do not have Python installed yet, it is highly recommended that you install the [Anaconda] (http://continuum.io/downloads) distribution of Python, which already has the above packages and more included. Make sure that you select the Python 2.7 installer and not the Python 3.x installer.