Skip to content

Dataset · zhihefang

Ultra-high-resolution images for text-to-image

zhihefang/UltraHR-100K

A dataset of ultra-high-resolution (UHR) images with captions for training high-detail image generation.

What's inside

The dataset compiles ultra-high-resolution images and English text descriptions for text-to-image tasks. Metadata indicates a volume of under 1,000 entries, although the title and description state the project is large-scale. The data is designed to address the shortage of quality UHR samples and refine fine details during generation.

What you can do with it

  • Fine-tuning text-to-image models (e.g., Flux or SDXL) for ultra-high-resolution image generation
  • Training LoRA adapters to enhance texture detail
  • Evaluating the quality of fine-detail synthesis in UHR scenarios
Tasks
Text → image
Languages
English
Size
up to 1K rows
Downloads this month
4.4K
Likes
37

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("zhihefang/UltraHR-100K")