watch the demo on youtube:

https://www.youtube.com/watch?v=Jgc8eNd_4Eg&t=57s

Deep-learning-Colorization-for-visual-media

Automatic colorization for grayscale (images/ videos) based on Deep-Learning using (Python & Tensorflow).

• A TensorFlow version of (Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic Image Colorization with Simultaneous Classification) http://hi.cs.waseda.ac.jp/~iizuka/projects/colorization/en/ Applied on Landscape images only(2000 images from MIT Places) data-set for leak of resources.

TODO :

• implement videos part
• apply and re-train the model using classification network
• train the model on more than one category.
• find all information needed in project documentation and presentation files from google drive link below. https://drive.google.com/drive/folders/0B7FFqObqrfTDOG5OSVZ4UUJ5S1U?usp=sharing

to test .exe version

https://drive.google.com/drive/folders/0B7FFqObqrfTDTWJMcE5UN3I3UjQ?usp=sharing

• note: it's trained only on simple landscape images(2000 image) so it will not give a good result if tried to test anothe category or complex images
• download .exe file attached from google drive and instruction file
• read insteraction to use .exe.txt to know how to run for first time
• Colorization trained model weights https://drive.google.com/file/d/1QzTMl8jHeowjzRP7DcfIQk4IpwVyTaGo/view?usp=sharing
• let me know if you faced any problems with loading the pre-trained weights

Notes

• project is still under maintenance and processing.
• you should be aware of change file pathes corrosponding to your data set locations when try to read or test images
• you should be aware of handling creating Model weight directory for first time to save trained model's weights every time you perform tarining.
• There is a single file containing the whole code as functions used temporarily for testing.

for any questions: [email protected]