Please read my blog to get a detailed insight for this project.
In this project, I am generating new human faces which probably does not exist in real life. It would be quite obvious for you that I will be using the generative adversarial network(GAN) for the task. I am using CelebA dataset for training the network. This dataset contains 2,00,000 images of well-known people.
Following are the generated faces after training.
We can generate more realistic images by making our network deeper, but it will take a lot of time to train the model. We can also modify our network to produce high-resolution images but again at the cost of training time.