• Publications
  • Influence
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
  • C. Ledig, Lucas Theis, +6 authors W. Shi
  • Computer Science, Mathematics
  • IEEE Conference on Computer Vision and Pattern…
  • 15 September 2016
SRGAN, a generative adversarial network (GAN) for image super-resolution (SR), is presented, to its knowledge, the first framework capable of inferring photo-realistic natural images for 4x upscaling factors and a perceptual loss function which consists of an adversarial loss and a content loss. Expand
Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network
This paper presents the first convolutional neural network capable of real-time SR of 1080p videos on a single K2 GPU and introduces an efficient sub-pixel convolution layer which learns an array of upscaling filters to upscale the final LR feature maps into the HR output. Expand
Lossy Image Compression with Compressive Autoencoders
It is shown that minimal changes to the loss are sufficient to train deep autoencoders competitive with JPEG 2000 and outperforming recently proposed approaches based on RNNs, and furthermore computationally efficient thanks to a sub-pixel architecture, which makes it suitable for high-resolution images. Expand
Bayesian Active Learning for Classification and Preference Learning
This work proposes an approach that expresses information gain in terms of predictive entropies, and applies this method to the Gaussian Process Classier (GPC), and makes minimal approximations to the full information theoretic objective. Expand
Amortised MAP Inference for Image Super-resolution
A novel neural network architecture is introduced that performs a projection to the affine subspace of valid SR solutions ensuring that the high resolution output of the network is always consistent with the low resolution input, and it is shown that the GAN based approach performs best on real image data. Expand
How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
This paper presents a critique of scheduled sampling, a state-of-the-art training method that contributed to the winning entry to the MSCOCO image captioning benchmark in 2015, and presents the first theoretical analysis that explains why adversarial training tends to produce samples with higher perceived quality. Expand
Faster gaze prediction with dense networks and Fisher pruning
Through a combination of knowledge distillation and Fisher pruning, this paper obtains much more runtime-efficient architectures for saliency prediction, achieving a 10x speedup for the same AUC performance as a state of the art network on the CAT2000 dataset. Expand
Variational Inference using Implicit Distributions
This paper provides a unifying review of existing algorithms establishing connections between variational autoencoders, adversarially learned inference, operator VI, GAN-based image reconstruction, and more, and provides a framework for building new algorithms. Expand
Optimally-Weighted Herding is Bayesian Quadrature
It is shown that sequential Bayesian quadrature can be viewed as a weighted version of kernel herding which achieves performance superior to any other weighted herding method. Expand
Collaborative Gaussian Processes for Preference Learning
A new model based on Gaussian processes for learning pair-wise preferences expressed by multiple users is presented which allows for supervised GP learning of user preferences with unsupervised dimensionality reduction for multi-user systems. Expand