Classification of images

Challenge: Creating new images

Neural

Net

Problem: How to train nets?

Task

+

GAN

noise ~

yes / no

generator sample

discriminator

data sample

data sample?

generator

How it works - "No Weenies allowed"

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

How it works - No Weenies allowed

GAN

noise

yes / no

generator sample

discriminator

data sample

data sample?

generator

GAN - Learning process

noise

yes / no

discriminator

data sample?

generator

GAN - Learning process

noise

yes / no

discriminator

data sample?

generator

GAN - Learning process

noise

yes / no

discriminator

data sample?

generator

Learning

process

Learning Progress

Reward

GAN Example

I. Problem - Trouble with counting

II. Problem - Trouble with perspective

III. Problem - Misunderstanding global structure

Outlook - Variational Autoencoder

Outlook - Adaptive Image Compression

Applications

References

• https://blog.openai.com/generative-models/
• http://gabgoh.github.io/ThoughtVectors/
• https://ishmaelbelghazi.github.io/ALI/
• http://kvfrans.com/variational-autoencoders-explained/
• https://arxiv.org/pdf/1701.00160.pdf
• https://research.googleblog.com/2016/11/zero-shot-translation-with-googles.html

• https://medium.com/@awjuliani/generative-adversarial-networks-explained-with-a-classic-spongebob-squarepants-episode-54deab2fce39

• https://arxiv.org/pdf/1705.05823.pdf

Recap - GAN

noise

yes / no

generator sample

discriminator

data sample

data sample?

generator

Recap - VAE

VAE with better colors

AE with better colors