YE-POP
YE-POP is an artistic image dataset derived and filtered from Laion-POP, with human validation and NSFW sorting to ensure quality.
11 chunks of 50,000 image URLs each, or about 550,000 images, URL format
Apache 2.0
Description
YE-POP is a dataset composed of carefully filtered and validated artistic images, extracted from Laion-pop. It contains over 550,000 image URLs divided into 11 chunks, with NSFW sorting and human verification to ensure the high quality of associated descriptions.
What is this dataset for?
- Train diffusion models for the generation of artistic images
- Improve automatic image captioning systems
- Evaluate and compare multimodal models on artistic images
Can it be enriched or improved?
Yes, it is possible to add additional annotations, such as artistic categories, styles, or source metadata. The collection of local images or the integration of royalty-free images would further enrich the dataset.
🔎 In summary
🧠 Recommended for
- Researchers in image generation
- Teams working on broadcasting and captioning
- AI art projects
🔧 Compatible tools
- Stable Diffusion
- SLAB
- Hugging Face Datasets
- OpenAI CLIP
- Diffusers
💡 Tip
Check the URLs before downloading, and complete with artistic metadata for added value.
Frequently Asked Questions
Does this dataset contain the images directly or only the URLs?
It only contains image URLs taken from Laion-pop, requiring images to be downloaded separately.
What is the estimated total size of the image dataset?
Approximately 550,000 images in total, divided into 11 chunks of 50,000 URLs each.
Can this dataset be used for commercial projects?
Yes, the Apache 2.0 license allows free commercial use under the terms of the license.
