Update README.md

# 📦 Freeze-Align Dataset

The **Freeze-Align Dataset** concept_coverage_laion_6m is a curated collection of high-quality image-text pairs designed to facilitate efficient multimodal alignment using frozen unimodal encoders. The dataset is curated from LAION400M by performing a concept-balanced collection of captions using caption-image-prototype similarity. The dataset supports the research presented in our CVPR 2025 paper "Harnessing Frozen Unimodal Encoders for Flexible Multimodal Alignment", enabling models to achieve CLIP-level performance with significantly reduced computational resources. The code and resources for curating this dataset are part of our code release.

## 📄 Paper

**Title:** Harnessing Frozen Unimodal Encoders for Flexible Multimodal Alignment
**Authors:** Mayug Maniparambil, Raiymbek Akshulakov, Yasser Abdelaziz Dahou Djilali, Sanath Narayan, Ankit Singh, Noel E. O'Connor
**Conference:** CVPR 2025
**Paper:** [arXiv:2409.19425](https://arxiv.org/abs/2409.19425)
**Code:** [GitHub Repository](https://github.com/mayug/freeze-align)

## 📊 Dataset Statistics

- **Total Samples:** 6,000,000 image-text pairs
- **Source:** Filtered from LAION-400M using semantic similarity measures
- **Image Resolution:** Variable; standardized during preprocessing
- **Text Language:** Primarily English
- **Data Format:** Parquet files with fields: `image_url`, `caption`, `embedding_vector`, `similarity_score`
- **License:** CC-BY 4.0

## 🧪 Usage

This dataset is intended for training and evaluating multimodal models that align visual and textual representations. It is particularly useful for research in:

- Multimodal representation learning
- Cross-modal retrieval
- Zero-shot image classification
- Efficient training with frozen encoders

To load the dataset using the Hugging Face `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("mayug/concept_coverage_laion_6m")
```

## 📂 Dataset Structure

Each entry in the dataset includes:
- `image_url`: URL to the image
- `caption`: Associated textual description
- `similarity`: Cosine similarity score between image and text embeddings
- `IMAGENET_CLASS`: CLASS out of approx 2700 classes assigned to the sample in the curation process. (Please refer to the paper for more info.)
- `SCORE`: The score that current sample belongs to the specified IMAGENET_CLASS

## 📬 Citation

If you use this dataset in your research, please cite our paper:

```bibtex
@inproceedings{maniparambil2025harnessing,
title={Harnessing Frozen Unimodal Encoders for Flexible Multimodal Alignment},
author={Maniparambil, Mayug and Akshulakov, Raiymbek and Djilali, Yasser Abdelaziz Dahou and Narayan, Sanath and Singh, Ankit and O'Connor, Noel E},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2025}
}
```

---

For more details and updates, please visit our [GitHub Repository](https://github.com/mayug/freeze-align).

README.md

@@ -1,3 +1,17 @@
----
-license: cc-by-4.0
----
+---
+license: cc-by-4.0
+task_categories:
+- zero-shot-classification
+- text-to-image
+- image-to-text
+language:
+- en
+tags:
+- image-caption
+- high-concept-coverage
+- laion-subset
+- 6M
+pretty_name: free-align-concept_covered_6M
+size_categories:
+- 1M<n<10M
+---