Antonio Torralba

TL;DR: This work proposes an approach for local relighting that trains a model without supervision of any novel image dataset by using synthetically generated image pairs from another model, including a stylespace-manipulated GAN.

TL;DR: In this article, a class-conditional generative adversarial network (GAN) is proposed to generate realistic and diverse images without using manually annotated class labels, instead, their model is conditional on labels automatically derived from clustering in the discriminator's feature space.

TL;DR: The authors leverage the feedback signal provided by the forward process and learn an iterative update model, where at each iteration, the neural network takes the feedback as input and outputs an update on the current estimation.

TL;DR: This work introduces a new approach to image forensics: placing physical refractive objects, which are called totems, into a scene so as to protect any photograph taken of that scene, and unscrambles the light rays passing through the totems by estimating their positions in the scene and using their known geometric and material properties.

TL;DR: This work presents a non-linear object detector called Exemplar Network that efficiently encodes the space of all possible mixture models, and offers a framework that generalizes recent exemplar-based object detection with monolithic detectors.