68 datasets found

FigureQA is a dataset for visual question answering, containing line plots, bar charts, pie plots, and dot line plots.

DVQA is a dataset for visual question answering, containing bar charts and natural language patterns.

The VizWiz-VQA dataset is a large-scale visual question answering dataset that consists of 4,000 images with 10 crowd-worker answers each.

The VQAv2 dataset, containing open-ended questions on 265k images, with 5.4 questions per image on average.

ST-VQA dataset consists of 23,038 images with 31,791 question-answer pairs.

TextVQA dataset consists of 28,418 images with 45,336 questions.

CARETS is a systematic test suite to measure consistency and robustness of modern VQA models through a series of six fine-grained capability tests.

The TallyQA dataset is a large-scale open-ended visual counting dataset, which is well-suited to study statistical shortcuts.

A diagnostic VQA dataset based on abstract objects that enables a faster and less biased evaluation of spatial reasoning behavior in VQA compared with the original GRiD-3D dataset.

A comprehensive, simplified diagnostic VQA dataset with abstract objects that shows similar behavior to the original GRiD-3D dataset when learned by the two established VQA...

The CLEVR-Humans dataset consists of 32,164 questions asked by humans, containing words and reasoning steps that were unseen in CLEVR.

The dataset is used for image captioning and visual question answering.

Visual question answering dataset

The VQA-CPv2 dataset is used to study the robustness of VQA methods against linguistic biases.

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 dataset for the SMART-101 challenge consists of 101 unique puzzles that require a mix of several elementary skills, including arithmetic, algebra, and spatial reasoning,...

The CLEVR-CoGenT dataset is a dataset for visual question answering, where the questions consist on comparing the position of two objects.

GQA-OOD is a benchmark dedicated to the out-of-domain VQA evaluation.

GQA is a new dataset for real-world visual reasoning and compositional question answering.

The High Quality Image Text Pairs (HQITP-134M) dataset consists of 134 million diverse and high-quality images paired with descriptive captions and titles.