Relation252K: 218 Image Editing Tasks for Visual Relation Transfer

Relation252K contains source-target image pairs across 218 distinct image editing tasks, released by handsomeWilliam in 2025. It serves as the evaluation set for the RelationAdapter model, focusing on the transfer of visual relations within Diffusion Transformers.

Use Cases

• Benchmarking image-to-image translation performance across 218 task categories
• Training Diffusion Transformers to identify and apply visual relations
• Testing content-aware image editing consistency between source and target pairs

Strengths

• Includes 218 distinct image editing tasks
• Permissive Apache 2.0 license
• Structured as source-target pairs for supervised relation learning

Limitations

• Lack of explicit column documentation in the provided metadata
• Size discrepancy between the 252K name and the 10K-100K repository tag

Provenance

Source

RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers (ArXiv 2506.02528)

Freshness

Last updated October 2025.