/Gray-to-RGB

Converting an gray scale image to colour image using autoencoder

Primary LanguageJupyter NotebookMIT LicenseMIT

Image Colorization using Autoencoders

Convert an gray scale image into rgb image using Convolutional Autoencoder

Requirements

pip3 install tensorflow-gpu==1.13.1

How to use

Download the flower dataset from here.Dataset contains 210 flower images.

  • Extract file :

tar -xvf filename

mv ./filename/flower_images ../../

  • Training

python main.py ./flower_images

Loss Graph

  • Graphs depicts Training loss Model_loss

Network Architecture

Architecture

Results

Result after 3000 epochs

  • First row contains ground truth gray scale images.
  • Second row contains ground truth rgb images.
  • Third row contains images generated from the network. Results

Conclusion

Convolutional Autoencoder can be used to convert gray scale images into rgb images.Encoder is able to effectively learn images representaion into latent vector and decoder is able to map learned image representaion into color image.