CNN-generated images are surprisingly easy to spot... for now

CVPR 2020 • Sheng-Yu Wang • Oliver Wang • Richard Zhang • Andrew Owens • Alexei A. Efros

In this work we ask whether it is possible to create a "universal" detector for telling apart real images from these generated by a CNN, regardless of architecture or dataset used. To test this, we collect a dataset consisting of fake images generated by 11 different CNN-based image generator models, chosen to span the space of commonly used architectures today (ProGAN, StyleGAN, BigGAN, CycleGAN, StarGAN, GauGAN, DeepFakes, cascaded refinement networks, implicit maximum likelihood estimation, second-order attention super-resolution, seeing-in-the-dark)... (read more)

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract

Results from the Paper

Edit

Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

METHOD	TYPE
PatchGAN	Discriminators
Tanh Activation	Activation Functions
Instance Normalization	Normalization
Sigmoid Activation	Activation Functions
GAN Least Squares Loss	Loss Functions
Cycle Consistency Loss	Loss Functions
CycleGAN	Generative Models
Softmax	Output Functions
ReLU	Activation Functions
Conditional Batch Normalization	Normalization
Residual Block	Skip Connection Blocks
TTUR	Optimization
GAN Hinge Loss	Loss Functions
Residual Connection	Skip Connections
Non-Local Operation	Image Feature Extractors
Non-Local Block	Image Model Blocks
Truncation Trick	Latent Variable Sampling
Linear Layer	Feedforward Networks
Dot-Product Attention	Attention Mechanisms
Projection Discriminator	Discriminators
Spectral Normalization	Normalization
Off-Diagonal Orthogonal Regularization	Regularization
Adam	Stochastic Optimization
Batch Normalization	Normalization
Early Stopping	Regularization
1x1 Convolution	Convolutions
SAGAN Self-Attention Module	Attention Modules
SAGAN	Generative Adversarial Networks
BigGAN	Generative Models
Convolution	Convolutions
Adaptive Instance Normalization	Normalization
R1 Regularization	Regularization
Leaky ReLU	Activation Functions
Dense Connections	Feedforward Networks
Feedforward Network	Feedforward Networks
StyleGAN	Generative Models

METHOD

TYPE

PatchGAN