Moondream2 COYO 5M Captions
Moondream2 COYO 5M Captions is a large dataset of images with their long captions automatically generated from alternative texts, selected according to criteria of aesthetic quality and similarity. Perfect for training text-to-image models.
5 million image-text pairs, 1.5 GB, image and text formats (captions)
CC-BY-SA 4.0
Description
The dataset Moondream2 COYO 5M Captions contains 5 million images associated with long captions generated from alternative descriptions (alt text). The images are filtered according to qualitative criteria, guaranteeing a varied and aesthetic corpus.
What is this dataset for?
- Train models for generating images from text
- Develop systems for understanding and automatically annotating images
- Improving the quality of long legends in multimodal applications
Can it be enriched or improved?
Yes, it is possible to add human annotations or additional metadata, and to refine the legends generated to improve the accuracy and diversity of the dataset.
🔎 In summary
🧠 Recommended for
- Computer vision researchers
- Multimodal AI
- Image generation
🔧 Compatible tools
- Hugging Face Transformers
- Diffusers
- PyTorch
- TensorFlow
💡 Tip
Use additional cleaning techniques to optimize legends according to the use case.
Frequently Asked Questions
What is the size of the Moondream2 COYO 5M Captions dataset?
The dataset contains approximately 5 million image-text pairs, with a total weight of 1.5 GB.
Under what license is this dataset available?
It is licensed under CC-BY-SA 4.0, which allows commercial use under the condition of sharing the same.
Can this dataset be used to train text-to-image models?
Yes, it is specially designed for fine-tuning models for generating images from text.
