https://github.com/tamohannes/image_colorization

Grayscal image to colorful transformation
https://github.com/tamohannes/image_colorization

cnn deep-learning keras neural-network tensorflow vgg16

Last synced: about 2 months ago
JSON representation

Grayscal image to colorful transformation

• Host: GitHub
• URL: https://github.com/tamohannes/image_colorization
• Owner: tamohannes
• Created: 2020-03-01T09:22:34.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2020-06-19T20:56:19.000Z (about 6 years ago)
• Last Synced: 2025-07-31T18:56:11.110Z (11 months ago)
• Topics: cnn, deep-learning, keras, neural-network, tensorflow, vgg16
• Language: Jupyter Notebook
• Homepage:
• Size: 6.74 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# image_colorization

## Grayscale image to colorful transformation - the aim is to get plausable colorful images which will "trick" the observer to give the feeling of a real RGB image

A VGG16 based CNN neaural network has been used, the schema is represented bellow:

![](model_plot.png)

Training sets are taken from Kaggle, practically any dataset could be used which consists of colorfull images (it's suggested to use a particullar class of images e.g. portrait, landscape)

Below are some results: (trainings are done by using two loss functions, MSE(in the middle) and Euclidean distance(on the right))

![](result1.png)

![](result2.png)

![](result3.png)

![](result4.png)

![](result5.png)

![](result6.png)

![](result7.png)