Colorizing Grayscale Images
Giving life (and color) to boring monochrome images.
Image colorization is a translation task that tries to estimate a plausible-looking colorized version of a given monochrome/grayscale image. Since there is a one-to-many mapping from grayscale to a possible colorized image, this is a highly undetermined problem.
I implemented the model as ResNet50 Autoencoder that translates an image from the Lab colorspace to the RGB colorspace in a supervised manner.
A couple of sample outputs from my model.
More details and code for this project are available here.