I have been recently working on a Voice Conversion system based on Convolutional Neural Networks: while the main idea behind the whole system wasn’t too complicated, I was getting pretty convincing results.
After giving myself a pat on the shoulder, I was about to start off a completely different project I had in mind, leaving my old project in a random folder on my busy desktop.
Then, suddenly, a lightning struck from the sky:
a Paper! An Academic Paper!
It really was an alluring idea, one that if conquered would provide immense satisfaction. Especially being in my situation.
I am an undergraduate, studying a subject that is completely unrelated to Machine Learning, even to Computer Science. I came to learn some Machine Learning basics guided by passion and only passion, without any academic perspectives on the whole field. I don’t know anything about how the academic world really works. …
We have all heard about image style transfer: extracting the style from a famous painting and applying it to another image is a task that has been achcieved with a number of different methods. Generative Adversarial Networks (GANs in short) are also being used on images for generation, image-to-image translation and more.
Climate Change is real.
And even though many scientists agree on the fact that we are already too late, people are just becoming conscious about this problem. And with the people comes politics, and with politics comes the money.
That’s why in the next years there will be a major push towards research in the Energy Sector, and Data Science is going to play a big role in this huge battle. Finding new patterns in the data is a clear path to obtaining powerful solutions for our energy-hungry world.
In this article, we will take a look at some problematic case-scenarios in which Machine Learning and data-driven techniques are being proven to offer great solutions, possibly making this field one of the main protagonists in the war against Climate Change. …
Generative Adversarial Networks, or GANs, have seen major success in the past years in the computer vision department. We are now able to generate highly realistic images in high definition thanks to recent advancements like StyleGAN from Nvidia and BigGAN from Google; often the generated or ‘fake’ images are completely indistinguishable from the real ones, defining how far GAN developement has really come.
A year ago I decided to begin my journey into the world of Generative Adversarial Networks, or GANs. I’ve always been intrigued by them since the beginning of my interest in Deep Learning, mainly for the incredible results that they could produce. When I think of the term Artificial Intelligence, GAN is one of the first words that come to my mind.
A number of recent studies explored some ways and techniques to generate High Definition images (1024x1024 pixels) using GANs (Generative Adversarial Networks). It’s incredibily surprising to see super realistic, HD images of human faces, animals and other things generated by an algorithm, especially remembering the first GAN images from just a few years back. We have gone from low quality, pixelated images to close-to-reality images in no time: this is a really clear proof of how quickly research in this field advances.
A Generative Adversarial Network is an extremely interesting deep neural network architecture able to generate new data (often images) that resembles the data given during training (or in mathematical terms, matches the same distribution).
Immediately after discovering GANs and how they work, I got intrigued. There is something special, maybe magical, about generating realistic looking images in an unsupervised manner. One area of GAN research that really caught my attention has been image-to-image translation: the ability to turn an image into another image keeping some sort of correspondence (for example turning a horse into a zebra or an apple into an orange). …