BigBiGAN

Introduced by Donahue et al. in Large Scale Adversarial Representation Learning

BigBiGAN is a type of BiGAN with a BigGAN image generator. The authors initially used ResNet as a baseline for the encoder $\mathcal{E}$ followed by a 4-layer MLP with skip connections, but they experimented with RevNets and found they outperformed with increased network width, so opted for this type of encoder for the final architecture.

Source: Large Scale Adversarial Representation Learning

Read Paper See Code

Papers

Paper	Code	Results	Date	Stars

Tasks

Task	Papers	Share
Image Classification	2	13.33%
Image Generation	2	13.33%
Classification	1	6.67%
Denoising	1	6.67%
Fine-Grained Image Classification	1	6.67%
Super-Resolution	1	6.67%
Saliency Detection	1	6.67%
Self-Supervised Learning	1	6.67%
Semantic Segmentation	1	6.67%

Usage Over Time

This feature is experimental; we are continuously improving our matching algorithm.

Components

Component	Type	Add Remove
Batch Normalization	Normalization
BigGAN	Generative Models
CReLU	Activation Functions
Feedforward Network	Feedforward Networks
Global Average Pooling	Pooling Operations
Linear Layer	Feedforward Networks
Residual Block	Skip Connection Blocks
RevNet	Convolutional Neural Networks
Softplus	Activation Functions

Categories

Add Remove

Generative Models

Generative Adversarial Networks

Self-Supervised Learning