Skip to content
Convolutional neural networks as an answer to image scaling issues
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
deep_filters
test_images
.gitignore
.theanorc
LICENSE
README.md
image_download.py
logging.conf
requirements.txt

README.md

Better image scaling with convolutional neural network

A convolutional neural network tested and trained in scaling up low-resolution images.

About

This is a trained and tested convolutional neural network based on Keras and Theano. Its purpose is to resolve a bad quality issue when scaling up a small, low-resolution image by big percentage. The network was trained on couple thousand images with approximately 5,000 images per epoch.The image download was also automatized for better efficiency.

More information here: http://build.sh/convolutional-neural-networks-as-an-answer-to-image-scaling-issues/

Network Architecture

The convolution layer has 150 9 x 9 filters with a 200 x 200 sized images being the input. After that comes the activation layer (RELU) followed by output layer. The used optimizer is Adam on default parameters.

Setup conda environment:

  • create your conda environment
  • install requirements from requirements.txt

Image Dowload:

In case you did not use Theano framework, amend and put .theanorc into your home directory

Network Learning:

  • python deep_filters/process-zoom2.py

Testing:

  • python deep_filters/process-zoom2.py --file image_to_enlarge.png
You can’t perform that action at this time.