metadata
license: openrail
task_categories:
- image-to-text
language:
- en
tags:
- Retrieval
size_categories:
- 1K<n<10K
Img2Text-Plaintext-Retrieval Dataset
Dataset Overview
The Img2Text-Plaintext-Retrieval dataset is designed for retrieving plaintext descriptions from corresponding algorithm images. This dataset consists of structured text, raw text, algorithm images, and metadata such as source URLs and filenames. It is suitable for tasks like OCR-based text retrieval, image-to-text learning, and document understanding.
Dataset Details
- Modality: Image, Text
- Format: Parquet
- Size: ~456MB
- Total Rows: 4,471
Features
| Column Name | Data Type | Description |
|---|---|---|
| structure | String | LaTeX or structured representation of the algorithm |
| text | String | Extracted textual representation of the algorithm |
| image | Image | Algorithm snapshot (from research papers) |
| download_url | String | URL of the source document |
| instance_name | String | Name of the instance (e.g., paper ID) |
Usage
This dataset is useful for:
- Algorithm text-image retrieval: Matching textual algorithm descriptions to corresponding images.
- OCR and text recognition: Evaluating OCR models for structured text extraction.
- Machine Learning research: Training deep learning models for image-to-text conversion.