Antonio Torralba

TL;DR: This work proposes a new contrastive loss function that is robust against noisy views and provides rigorous theoretical justifications by showing connections to robust symmetric losses for noisy binary classification and by establishing a new Contrastive bound for mutual information maximization based on the Wasserstein distance measure.

TL;DR: An oracle neural-symbolic framework named Compositional Physics Learner (CPL), combining visual perception, physical property learning, dynamic prediction, and symbolic execution into a unified framework that can effectively identify objects’ physical properties from their interactions and predict their dynamics to answer questions.

TL;DR: This work identifies and study two types of “accidental” images that can be formed in scenes, one of which is an accidental pinhole camera image, and the other is “inverse”Pinhole camera images, formed by subtracting an image with a small Occluder present from a reference image without the occluder.

TL;DR: In this article, computer vision can be used to infer a person's body mass index (BMI) from social media images, which can have profound implications on their life, ranging from mental health, to longevity, to financial income.

TL;DR: Wang et al. as mentioned in this paper proposed 3D-interpreter networks (3D-INN) to estimate 2D keypoint heatmaps and 3D object skeletons and poses from real images and synthetic 3D shapes.