Omarrran's picture
Update README.md
2ece2f7 verified
metadata
license: apache-2.0

Vision-Language Pairs Dataset

This dataset contains metadata about image-text pairs from various popular vision-language datasets.

Contents

  • vision_language_data/all_vision_language_images.csv: Combined metadata for all images (75629 records)
  • vision_language_data/all_vision_language_captions.csv: Combined captions for all images (86676 records)
  • dataset_statistics.csv: Summary statistics for each dataset
  • category_distribution.csv: Distribution of image categories across datasets
  • caption_length_distribution.csv: Distribution of caption lengths
  • caption_style_distribution.csv: Distribution of caption styles
  • category_caption_statistics.csv: Caption statistics by category
  • vision_language_catalog.json: Searchable catalog with sample image-caption pairs

Datasets Included

  • COCO (Common Objects in Context): COCO is a large-scale object detection, segmentation, and captioning dataset with multiple captions per image. (123287 images)
  • Flickr30K (Flickr 30,000 Images): Flickr30K contains images collected from Flickr with 5 reference captions per image provided by human annotators. (31783 images)
  • Visual Genome (Visual Genome): Visual Genome connects structured image concepts to language with detailed region descriptions and question-answer pairs. (108077 images)
  • Conceptual Captions (Conceptual Captions): Conceptual Captions is a large-scale dataset of image-caption pairs harvested from the web and automatically filtered. (3300000 images)
  • CC3M (Conceptual 3 Million): CC3M is a dataset of 3 million image-text pairs collected from the web, useful for vision-language pretraining. (3000000 images)
  • SBU Captions (SBU Captioned Photo Dataset): The SBU dataset consists of 1 million images with associated captions collected from Flickr. (1000000 images)

Fields Description

Images Table

  • image_id: Unique identifier for the image
  • dataset: Source dataset name
  • image_url: URL to the image (simulated)
  • primary_category: Main content category
  • width: Image width in pixels
  • height: Image height in pixels
  • aspect_ratio: Width divided by height
  • caption_count: Number of captions for this image
  • license: License under which the image is available

Captions Table

  • caption_id: Unique identifier for the caption
  • image_id: ID of the associated image
  • dataset: Source dataset name
  • text: Caption text
  • language: Caption language (default: en)
  • style: Caption style (descriptive, short, or detailed)
  • length: Number of characters in the caption
  • word_count: Number of words in the caption

Usage Examples

This metadata can be used for:

  1. Analyzing the composition of vision-language datasets
  2. Comparing caption characteristics across different datasets
  3. Training and evaluating image captioning models
  4. Studying linguistic patterns in image descriptions
  5. Developing multimodal AI systems

Data Generation Note

This dataset contains synthetic metadata that represents the structure and characteristics of actual vision-language pair collections, but the specific image and caption details are generated for demonstration purposes.

Created: 2025-04-26

Note:

All files are packaged into a ZIP archive called vision_language_pairs_data.zip for easy download, with expected size in the 150-200MB range, making it suitable for research and educational purposes.