Race change in facial image of a person

Race-change-in-facial-images-using-Cycle-Generative-Adversarial-Networks

The model uses Cycle Generative Adversarial Networks (latest invention in Deep Learning) to change the race of the person in the facial images of people.

Image-to-image translation is a class of vision and graphics problems where the goal is to learn the mapping between an input image and an output image using a training set of aligned image pairs. However, for many tasks, paired training data will not be available and this is the main obstacle for the application of image to image translation in different domains.The latest invention in deep learning called as Cycle Generative Adversarial Networks destroys this barrier using a new type of loss in neural networks called as Cycle Loss.

Cycle GAN has many applications such as generation of monet paintings ,novel music and medical imaging.For this hackathon , I developed a Cycle GAN from scratch to change the race of the person in the facial images of the people, without losing the original characteristics and properties of face fed as input to the model.My model differentiates itself from the existing machine learning models for style and race transfer by using a totally unpaired image dataset.

No of Neural networks used : 4.

Architecture of the model : alt text