RotNIST

Rotated MNIST dataset This project was designed to create rotations of the popular MNIST digits(http://yann.lecun.com/exdb/mnist/).

The generative process used to generate the datasets is the following:

1. Pick a sample from the MNIST digit recognition dataset;
2. Create a rotated version of the sample. The digits were rotated by an angle generated randomly between -\pi and \pi radians.
3. Go back to 1 until enough samples(by default 4, can be passed as argument) are generated.

INSTALLATION OF REQUIRED TOOLS

1. Tensorflow

Refer to the following link https://www.tensorflow.org/install/install_sources. Tensorflow is used as backend for Keras. The link contains installation instructions with and without gpu support

2.Skimage

Scikit-image is an image processing toolbox for SciPy. It is used for loading,saving and applying various transformations like color to gray and gray to color on images.

• Refer following link for installation instructions http://scikit-image.org/docs/dev/install.html

3. numpy

No introductions required!!!

• To install numpy

  sudo pip install numpy

4. PIL

Pillow is the friendly PIL fork by Alex Clark and Contributors. PIL is the Python Imaging Library by Fredrik Lundh and Contributors.

• To install PIL

  sudo pip install pillow

5. Scipy

SciPy (pronounced “Sigh Pie”) is a Python-based ecosystem of open-source software for mathematics, science, and engineering.

• To install Scipy

  sudo pip install scipy

6. SIX

Six is a Python 2 and 3 compatibility library. It provides utility functions for smoothing over the differences between the Python versions with the goal of writing Python code that is compatible on both Python versions.

• To install SIX

  sudo pip install six

Quick start

Run the following commands in Terminal.

git clone https://github.com/ChaitanyaBaweja/RotNIST.git

cd RotNIST

python main.py

python images.py

• main.py Contains the code to save rotated images in 4D numpy arrays and return them for other codes, to be used as auxiliary file for networks
• images.py Contains the code to save rotated images as .jpg files

Soon to come

• Wrapper function to put all these together and pass arguments
• Faster code implementation
• Results of this data set on Capsule networks

Contact Details:

Chaitanya Baweja : [email protected]