Google Code-in

Background Remover

This is a very common project, but also very much required by design professionals worldwide. A background remover tool identifies the subject from the background, and removes the background.

Separating foreground from the background image can be quite the task for all sorts of images, you may focus your work only on human portraits. You have to separate the human image layer from the background layer. This is a task of deep learning.

You can use the Tiramisu model to train your data set. The main challenge here is to acquire and train the dataset. You should use a variety of images, which have a marked distinction between the foreground and the background, so that your model can learn how to separate them.

You can take help from the following tutorial, it is detailed and beautifully explains some concepts of deep learning : https://towardsdatascience.com/background-removal-with-deep-learning-c4f2104b3157

You have to upload the code, and the data-set used on your GitHub profile, and then send the link of your repository for review.