[Pycon] [new paper] "Michele De Simoni" - AbracaGAN: introduction to the almost magical world of GANs

[info a pycon.it]

Title: AbracaGAN: introduction to the almost magical world of GANs
Duration: 60 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: # AbracaGAN: Introduction to the almost magical world of GANs

**Requires:** entry-level Python and basic Deep Learning knowledge (Convolutional Layers, Gradient Descent)

Generative Adversarial Networks (GANs for brevity) are one of the hottest topics in the Deep Learning world, and as is often the case, they can seem magical black boxes to the uninitiated or a cool research topic with no real-world applications to the eyes of the business people. In this talk, I will try and show you why both views are (almost) entirely wrong.

After briefly introducing the core theory behind this family of models, we will go over some significant research milestones and real-world applications.

Having a better understanding of what a GAN is and how it can solve real-world problems, the talk will cover how to code one using Python + Keras + TensorFlow 2.0 and the few steps required to deliver it quickly to production using Google Cloud ML Engine.

The Talk will feature slides, materials under the form of Jupyter Notebooks on Colab will be made available before the session.


Tags: [u'ComputerVision', u'neural network', u'gans', u'tensorflow', u'deep learning', u'generative-models', u'pydata']