OmniConsistency Stylized Image Pair Dataset

OmniConsistency is a large-scale multi-style image pair dataset.OmniConsistency: Learning Style-Agnostic Consistency from Paired Stylization Data", focusing on image stylization and cross-modal consistency learning, aims to provide standardized resources for image generation, style transfer and multimodal model training.

The dataset covers 22 different art styles, including cartoons (3D_Chibi, American_Cartoon), oil paintings (Oil_Painting, Van_Gogh), traditional art (Chinese_Ink, Paper_Cutting), and pixel art (Pixel), to meet diverse creative needs.

Each style consists of aligned image pairs:

• src: Original image (such as a photograph or line sketch)
• tar: Stylized image
• prompt: Descriptive text representing the intended art style

This dataset is suitable for the following tasks:

• Style Transfer
• Image to Image Generation
• Conditional generation with prompts
• Consistency Learning