PerSeg is a dataset for personalized segmentation. The raw images are collect from the training data of subject driven diffusion models: DreamBooth, Textual Inversion, and Custom Diffusion. PerSeg contains 40 objects of various categories in total, including daily necessities, animals, and buildings. Contextualized in different poses or scenes, each object is related with 5∼7 images with our annotated masks.
Source: Personalize Segment Anything Model with One Shot
Image Source: Personalize Segment Anything Model with One Shot
Variants: PerSeg
This dataset is used in 1 benchmark:
| Task | Model | Paper | Date |
|---|---|---|---|
| Personalized Segmentation | P^2SAM | Part-aware Personalized Segment Anything Model … | 2024-03-08 |
| Personalized Segmentation | PerSAM-F | Personalize Segment Anything Model with … | 2023-05-04 |
| Personalized Segmentation | PerSAM | Personalize Segment Anything Model with … | 2023-05-04 |
| Personalized Segmentation | Painter | Images Speak in Images: A … | 2022-12-05 |
| Personalized Segmentation | Visual Prompting | Visual Prompting via Image Inpainting | 2022-09-01 |
Recent papers with results on this dataset: