AI_SDXL_inpainting_with_GroundedSAM2

상위 문서로 이동 : AI Wiki

AI_SDXL_inpainting_with_GroundedSAM2

• The SDXL Inpainting model edits or fills parts of an image based on a mask and a text prompt.

Model Test

• Origin Image

• Mask Image We used Grounded SAM2 for masking to precisely define the areas for SDXL Inpainting. Grouded with SAM2

Example Output

1	2	3

• These images were generated using different prompts to explore various outputs.

Issue

• The mask images are a bit jagged, which results in less clean inpainting outputs.
• Imprecise masking of the 'desk' and 'deskmat' appears to cause issues in the final output.

Next Steps

• Generate a mask image that excludes the desk and deskmat from the inpainting area.

Reference

diffusers/stable-diffusion-xl-1.0-inpainting-0.1