Have you ever think how do we imagine? We got some signals like sounds, images, or perception, and in our mind appear different images. Do you mind, neural networks can do the same?

GAN (Generative Adversarial Network) - is a kind of AI that can "draw" or "imagine" different images according to the specific input signals. There are many types and variations, but now the word is about DCGAN (Deep Convolutional GAN). We implemented it to the Fashion Dataset with specific attributes for each image (e.g. color, texture, gender, etc).

So we have made a model that can generate clothes by description! It might be totally new unseen fantasy of your mind written in text form and translated into the real image of clothes.

Stay with us and you will see not only 128x128 images but good quality too! It will be our next stage.