ImgEdit is a large-scale, high-quality image-editing dataset comprising 1.2 million carefully curated edit pairs, which contain both novel and complex single-turn edits, as well as challenging multi-turn tasks.
Variants: ImgEdit-Data
This dataset is used in 1 benchmark:
| Task | Model | Paper | Date |
|---|---|---|---|
| Image Editing | BAGEL-NHR-EDIT | NoHumansRequired: Autonomous High-Quality Image Editing … | 2025-07-18 |
| Image Editing | UniWorld-V1 | UniWorld-V1: High-Resolution Semantic Encoders for … | 2025-06-03 |
| Image Editing | BAGEL | Emerging Properties in Unified Multimodal … | 2025-05-20 |
| Image Editing | Step1X-Edit | Step1X-Edit: A Practical Framework for … | 2025-04-24 |
| Image Editing | AnyEdit | AnyEdit: Edit Any Knowledge Encoded … | 2025-02-08 |
| Image Editing | UltraEdit | UltraEdit: Instruction-based Fine-Grained Image Editing … | 2024-07-07 |
| Image Editing | ICEdit | In-Context Editing: Learning Knowledge from … | 2024-06-17 |
| Image Editing | MagicBrush | MagicBrush: A Manually Annotated Dataset … | 2023-06-16 |
| Image Editing | Instruct-Pix2Pix | InstructPix2Pix: Learning to Follow Image … | 2022-11-17 |
Recent papers with results on this dataset: