Python is an opensource language that has a wide community, and nowadays the biggest companies also create frameworks, libraries with Python and open- sourced them. Tensorflow is the most used library in Deep Learning by researchers and there are many examples of various fields like Computer Vision, Natural Language Processing, Signal Processing.
Nowadays, Generative Adversarial Networks a.k.a. GAN collect nearly all interests on it by the Computer Vision experts. There are diverse applications like image colorization, image generation from random numbers, computer game character creation, face frontalization, face alignment, 2D to 3D image transfer, style transfer and so on.
In this talk, we are going to talk about GANs and the implementation details on Tensorflow which is backed by Google and has the power of either work on CPU and GPU.
The implementation of GANs can be divided into 2 parts. One is called generator and other is called discriminator. In this talk, the differences between the discriminator and generator also are mentioned.
Besides, some various architecture of GANs like PGGAN, DCGAN, STARGAN, architectures showed in Tensorflow code, which of these are not already implemented in Tensorflow by the date now.
The session will be finished with showing some examples of outputs in face generation, room generation, and also live demo of the style transfer implementation.
in __on sabato 21 aprile at 15:30 **See schedule**