Aesthetic 4K
The Aesthetic-4K dataset is dedicated to the generation of ultra-high resolution images. It contains carefully selected images and captions generated automatically by Gpt-4o. Manual filtering eliminated blurry or poor quality images, ensuring an excellent quality corpus for training advanced models.
Description
Aesthetic-4K is a dataset of carefully selected ultra-high resolution images, with captions generated by Gpt-4o. The dataset was cleaned manually to ensure image quality by eliminating blurs, focus issues, and textual inconsistencies.
What is this dataset for?
- Train ultra-high resolution image generation models
- Test and evaluate the synthesis of detailed and aesthetic images
- Improve the quality and consistency of automatic captions associated with images
Can it be enriched or improved?
The dataset can be enriched by adding new ultra-high resolution images or by improving automatic annotations via other language models or human annotations.
🔎 In summary
🧠 Recommended for
- Computer vision researchers
- Broadcast model developers
- Artistic AI projects
🔧 Compatible tools
- Diffusers
- PyTorch
- TensorFlow
- High-resolution image processing tools
💡 Tip
Take advantage of GPT-4o legends to guide fine-tuning on controlled image generation tasks.
Frequently Asked Questions
What is the size of the Aesthetic-4K dataset?
Approximately 2,700 ultra-high resolution images, totaling 10 GB in parquet format.
What are the characteristics of annotations?
The images are accompanied by captions generated automatically by GPT-4o, filtered for quality.
What license does this dataset cover?
The dataset is under the MIT license, free to use, including commercial.
