15 datasets found

MathVista is a benchmark for evaluating mathematical reasoning in visual contexts.

Inferring and executing programs for visual reasoning.

A dataset and architecture for visual reasoning with a working memory.

The VQA-CP dataset is a split of the VQA dataset, designed to test generalization skills across changes in the answer distribution between the training and the test sets.

The CLEVR-ART dataset is a novel dataset consisting of ART problems rendered using more realistic 3D shapes, based on the CLEVR dataset.

The Synthetic Visual Reasoning Test (SVRT) dataset consists of 23 binary classification tasks, each defined by a particular configuration of relations. The tasks can be broadly...

The Abstract Reasoning Tasks (ART) dataset was proposed by Webb et al. [51], consisting of four visual reasoning tasks, each defined by a different abstract rule. In the...

Human visual reasoning is characterized by an ability to identify abstract patterns from only a small number of examples, and to systematically generalize those patterns to...

A benchmark for evaluating task compositionality and long-horizon tasks through object manipulation, with language serving as the mechanism for goal specification.

The CLEVR dataset is a dataset for visual question answering, where each image is annotated with a question.

The VTT dataset is a collection of instructional videos with annotated transformations.

The dataset used in the paper is a set of sequential vision-and-language tasks, where each task consists of an image and a text input.

A collection of heterogeneous visual reasoning data sets and an interesting benchmark for two reasons: First, visual reasoning programs tend to be large (in current program...

The dataset used in the paper is a visual reasoning task using Graph Neural Networks (GNNs) and Recurrent Independent Mechanisms (RIMs). The dataset consists of 8 Atari games...

The Visual Genome dataset is a large-scale visual question answering dataset, containing 1.5 million images, each with 15-30 annotated entities, attributes, and relationships.