This software uses convolutional neural network and aims to generate a ground truth set in the smallest time possible with a minimum number of human interactions.

#Features

• Import a folder of images to classify
• Create an aribtrary number of sets from images
• Convolutional neural network to train on sets
• Move items between sets once classified
• Save\Load sets of images
• View reports on how the sets and how the network is performing
• Change various settings such as the number of iterations or learning rate

Several libraries are required to run the software, and are outlined below:

• Python 3.6 - https://www.python.org/downloads/
• Tensorflow (pip recommended) - https://www.tensorflow.org/install/
• PyQt5 (pip recommended) - http://pyqt.sourceforge.net/Docs/PyQt5/installation.html
• numpy + matplotlib (pip recommended) - https://www.scipy.org/install.html
• PIL - https://pillow.readthedocs.io/en/latest/installation.html

Then run the following to start the application:

$ cd Framework
$ python __init__.py