Lecture 13. Generative Models

 

Overview 

- Unsupervised Learning

- Generative Models

• PixelRNN and PixelCNN
• Variational Autoencoders (VAE) 
• Generative Adversarial Networks (GAN)

Classification : Input : Image Output : Text (Label)

Object Detection : Input : Image Output : Bounding Boxes of instances

Semantic Segmentation (having label for every pixel) : ?

Image Captioning : Input : Image Output : Caption (form of natural language) 

 

???

K-means clustering 

Principal Component Analysis (Dimensionality reduction) : 설명~~

Autoencoders (Feature Learning) : 설명~~

2-d density estimation 

 

 

 

So why do we care about Generative Models? 

able to create realistic samples from the data distributions

can get completely new generated samples