/Colorizing-Grayscale-Images

Utilize deep learning models to automatically colorize grayscale images

Primary LanguagePythonApache License 2.0Apache-2.0

Colorizing grayscale images

WHAT?

Convert grayscale image to a colored image using different deep learning techniques.

HOW?

We use 3 different models to try and colorize the grayscale image.

  1. Deep Koalarization [1]
  2. Inception-VGG AutoEncoder
  3. VGG AutoEncoder
  4. GAN (Experimental)

SETUP

DATA PREP

This project uses Microsoft COCO Datset[2]. This project uses 2017 train, validation and test images. But, any year data should work if retraining the model.

Place the images in ./data/train, ./data/validation and ./data/test folders.

INSTALLING DEPENDENCIES

  1. Install Python 3.6
  2. Install virtualenv
  3. Clone this repo
  4. cd into the repo
  5. Create a virtual environment
  6. Run pip install -r requirements.txt (Use requirements-gpu.txt if using a GPU)

LICENSE

This project is licensed under Apache License 2.0. The terms of the license can be found in LICENSE.

REFERENCES

  1. Baldassarre, Federico, Diego González Morín, and Lucas Rodés-Guirao. "Deep Koalarization: Image Colorization using CNNs and Inception-ResNet-v2." arXiv preprint arXiv:1712.03400 (2017).
  2. Lin, Tsung-Yi, et al. "Microsoft coco: Common objects in context." European conference on computer vision. Springer, Cham, 2014.