LearnKeras

This repository contains some code for you to get started with keras with a few simple datasets.

Compatibility

• This code runs on Python 3.5 and Keras 2.0.4 and has been tested on Ubuntu 16.04.

Intro

• The folder Intro contains the keras implementation for the analysis of the pima-indians-diabetes dataset.

• It involves the prediction of a binary output variable using 8 input variables.

• The code learnkeras1.py steps you through the training process and also shows you how to save your model checkpoint.

• The code learnkeras2.py shows you how to load the above trained model and find accuracy.

CIFAR-10

• The folder CIFAR-10 contains the code for classifying images into 10 categories of the famous CIFAR-10 dataset.

• I have used a simple model architecture for easy understanding of code.

• The code can be run in train or test mode.

• To run in train mode, type python cifar10.py --mode train. This trains the model and saves the checkpoint in the folder. It will also make predictions on test data and print the accuracy.

• To run in test mode, type python cifar10.py --mode test. This will output the test accuracy using weights from the checkpoint saved.

• Download the checkpoint for 10 epochs here.