---
license: cc-by-nc-4.0
task_categories:
- image-to-text
- text-retrieval
- text-to-image
language:
- en
pretty_name: LONG-DCI
size_categories:
- 1K<n<10K
---

## News  
We are excited to introduce **Long-DCI**, a new benchmark for long-caption image retrieval, adapted from the recently released Dense Captioning Images [DCI](https://github.com/facebookresearch/DCI) dataset (Urbanek et al., 2024). **Long-DCI** comprises 7,805 human-annotated image-caption pairs, with captions averaging 200 tokens per image.

---

### Long-DCI Dataset Card  

#### **Dataset Details**  
- **Dataset Type**: Long-DCI includes 7,805 image-caption pairs.
- **Download Datasets**: To download the DCI dataset, please follow the instructions provided in the [DCI GitHub repository](https://github.com/facebookresearch/DCI?tab=readme-ov-file#dataset-download). Once downloaded, you can use our CSV file for evaluation purposes.
- **More Information**:  
  - [Paper](https://arxiv.org/pdf/2410.10034)  
  - [Code](https://github.com/ivonajdenkoska/tulip)  
- **License**: Attribution-NonCommercial 4.0 International, in compliance with MetaAI’s policy.  

#### **Intended Use**  
- **Primary Purpose**: Long-DCI is designed for research on cross-modal retrieval.  
- **Target Audience**: This dataset is tailored for researchers and enthusiasts in computer vision, natural language processing, machine learning, and artificial intelligence.