Anything2Real Demo: Paired Reference and Target Images for Style Transfer

A dataset containing paired reference and target image sets, each accompanied by a text caption describing the transformation. It was created by author lrzjason and last updated on December 12, -2025. The dataset is designed for tasks involving image editing and controlled generation.

Use Cases

• Training models for style transfer based on paired reference and target images.
• Developing instruction-following image generation systems based on text captions describing transformations.
• Benchmarking controlled image transformation algorithms based on the provided image pairs.
• Fine-tuning image editing models based on the abstract image generation process mentioned.

Strengths

• Dataset structure is explicitly defined for paired reference-target image sets.
• Includes a text caption for each image pair, providing a multimodal signal.
• Last update timestamp (2025-12-12 14:49:35) is provided.

Limitations

• Row count, file formats, and column definitions are unknown, limiting suitability assessment.
• License information is not provided, which may restrict usage.
• The dataset is described as a 'demo', which may indicate a limited scale.

Provenance

Source

huggingface

Collection Method

Likely generated using a specific LoRA model (tori29umai/QwenImageEdit2509_LoRA/QIE_image2rakugaki_V1_dim4_1e-3-000012.safetensors) as mentioned in the description.

Time Range

null

Freshness

Last updated 2025-12-12 14:49:35.

Geography

null