Vision-Language Pre-training - Groups

Open Images

The Open Images dataset is a large-scale image dataset with a wide range of images, including but not limited to, street scenes, indoor scenes, and outdoor scenes.

Dataset
JSON

COCO 5K

The dataset used in the paper for unpaired vision-language pre-training via cross-modal CutMix.

Dataset
JSON

Conceptual Captions 3M

The Conceptual Captions 3M dataset is a large-scale image-text dataset used for vision-language pre-training.

Dataset
JSON

EPIC: Leveraging Per Image-Token Consistency for Vision-Language Pre-training

The proposed EPIC method is a pre-training approach that leverages more text tokens for learning vision-language associations.

Dataset
JSON

MixGen: A New Multi-Modal Data Augmentation

MixGen: a joint data augmentation for vision-language representation learning to further improve data efﬁciency.

Dataset
JSON

BookCorpus

The dataset used in this paper for unsupervised sentence representation learning, consisting of paragraphs from unlabeled text.

Dataset
JSON

COCO Captions

Object detection is a fundamental task in computer vision, requiring large annotated datasets that are difficult to collect.

Dataset
JSON

ALIGN

Scaling up visual and vision-language representation learning with noisy text supervision.

Dataset
JSON

Visual Genome

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.

Dataset
JSON

9 datasets found