DCGAN_Discriminator: DCGAN Discriminator

In mneunhoe/RGAN: Generative Adversarial Nets (GAN) in R

DCGAN Discriminator

Description

Provides a torch::nn_module with a simple deep convolutional neural net architecture, for use as the default architecture for image data in RGAN. Architecture inspired by: https://pytorch.org/tutorials/beginner/dcgan_faces_tutorial.html

Usage

DCGAN_Discriminator(
  number_channels = 3,
  ndf = 64,
  dropout_rate = 0.5,
  sigmoid = FALSE
)

Arguments

number_channels	The number of channels in the image (RGB is 3 channels)
ndf	The number of feature maps in discriminator
dropout_rate	The dropout rate for each hidden layer
sigmoid	Switch between a sigmoid and linear output layer (the sigmoid is needed for the original GAN value function)

Value

A torch::nn_module for the DCGAN Discriminator

mneunhoe/RGAN documentation built on Aug. 27, 2023, 7:57 a.m.