Residual Encoder Network for Colorization

Overview

This is a Tensorflow implementation of the Residual Encoder Network based on Automatic Colorization and the pre-trained VGG16 model from https://github.com/machrisaa/tensorflow-vgg

Structure

• config.py: config variables like learning rate, batch size and so on
• image_helper.py: all functions related to image manipulation
• read_input.py: input related functions
• residual_encoder.py: the residual encoder model
• batchnorm.py: batch normalization based on this
• common.py: the common part for training and testing, mainly the work flow for this model
• train.py: train the residual encoder model using Tensorflow built-in GradientDescentOptimizer
• test.py: test your own images and save the output images

Tensorflow graph

How to use

• First please download pre-trained VGG16 model vgg16.npy to vgg folder

• Use pre-trained residual encoder model

  • Model can be downloaded here
  • Unzip all files to model_path (you can change this path in config.py)

• Train your own model

  1. Change the learning rate, batch size and training_iters accordingly
  2. Change train_dir to your directory that contains all your training jpg images
  3. Run python train.py

• Test

  1. Change test_dir to your directory that contains all your testing jpg images
  2. Run python test.py

References

• Automatic Colorization
• pavelgonchar/colornet
• pretrained VGG16 npy file

License

GNU GPL 3.0 for personal or research use. COMMERCIAL USE PROHIBITED.