Bing-su/adetailer
Auto detecting, masking and inpainting with detection model.
What it solves
It automates the process of masking and inpainting specific objects (such as faces, hands, or people) in images generated by Stable Diffusion, removing the need for manual masking to fix details.
How it works
The tool operates in a three-step pipeline:
- Image Generation: An initial image is created.
- Detection: A detection model (such as YOLOv8 or MediaPipe) identifies the target object and generates a mask image.
- Inpainting: The system uses the original image and the generated mask to perform inpainting, refining the details of the detected object.
Who it’s for
Users of the AUTOMATIC1111 Stable Diffusion web UI who want to improve the quality of specific image elements without manual effort.
Highlights
- Automated Masking: Uses detection models to automatically find and mask targets.
- Flexible Detection: Supports various models for faces, hands, and people, as well as custom Ultralytics YOLO models.
- ControlNet Integration: Compatible with ControlNet models like inpaint, scribble, lineart, and openpose to guide the inpainting process.
- Customizable Masks: Offers options for mask erosion, dilation, offsets, and confidence thresholds to fine-tune the detection area.
Related
- Project
- Project
- Project
- Dispatch
- Project