README.md

Quick Segment for Stable Diffusion WebUI

Simplified fork focusing on quick segmentation using SAM (Segment Anything Model) and GroundingDINO.

This extension provides an easy-to-use interface for detecting and masking objects, body parts, clothing, and faces using AI-powered segmentation. Perfect for creating masks to use with img2img inpainting or other image editing workflows.

Features

• Quick Object Detection - One-click detection of:

  • Full body, upper/lower body
  • Face and facial features
  • Hands and feet
  • Clothing items (shirts, pants, dresses, etc.)
  • Background
  • Custom object detection via text prompts

• AI-Powered Segmentation - Uses state-of-the-art models:

  • SAM (Segment Anything Model) for precise mask generation
  • GroundingDINO for text-based object detection
  • Multiple SAM model options (standard, HQ, mobile)

• Simple Workflow:

  1. Upload your image
  2. Select a preset (or enter custom prompt)
  3. Click "Generate Mask"
  4. Select the best mask from the results
  5. Click "Send to Inpaint Upload" to start editing immediately

• Mask Customization:

  • Adjustable detection threshold
  • Mask expansion/dilation
  • Combined mask option for multiple detections
  • Invert mask option (great for background replacement)

How to Use

1. Open the extension in txt2img or img2img tabs (look for "Quick Segment" accordion)
2. Select SAM model from the dropdown (will auto-download if not present)
3. Upload your image
4. Choose a preset from the dropdown:
   • Full Body, Upper/Lower Body
   • Face, Eyes, Nose, Mouth, Ears
   • Hands, Feet
   • Shirts, Pants, Dress, Shoes
   • Or select "Custom" to enter your own text prompt
5. Adjust settings (optional):
   • Detection Threshold: Lower = more detections, Higher = stricter
   • Mask Expansion: Expand mask edges (5-15 pixels recommended)
   • Also Show Combined Mask: Adds an option to combine all detected regions
6. Click "Generate Mask" and wait for results
7. View results in the gallery:
   • Each row shows: [Preview | Mask | Cutout] for each detection
8. Send to Inpaint:
   • Select the best mask using the radio buttons below the gallery
   • (Optional) Check "Invert Mask" if you want to replace the background instead of the subject
   • Click "📋 Send to Inpaint Upload"
   • The extension will automatically switch you to the Inpaint Upload tab with your image and mask ready to go!

Installation

cd ${sd-webui}/extensions
git clone https://github.com/dmikey/sd-webui-quick-segment

Or install via the WebUI Extensions tab.

Choose one or more of the models below and put them to ${sd-webui}/models/sam or ${sd-webui-quick-segment}/models/sam. Do not change model name, otherwise this extension may fail due to a bug inside segment anything.

Supported SAM Models

1. SAM from Meta AI.

   • 2.56GB sam_vit_h
   • 1.25GB sam_vit_l
   • 375MB sam_vit_b

2. SAM-HQ from SysCV.

   • 2.57GB sam_hq_vit_h
   • 1.25GB sam_hq_vit_l
   • 379MB sam_hq_vit_b

3. MobileSAM from Kyung Hee University.

   • 39MB mobile_sam

GroundingDINO models will be automatically downloaded on first use.

Troubleshooting

GroundingDINO Installation Issues:

• This extension uses local GroundingDINO by default to avoid C++/CUDA compilation issues
• If you encounter errors, go to Settings → Quick Segment → Enable "Use local groundingdino"
• Models are auto-downloaded from HuggingFace on first use

MPS/Apple Silicon Issues:

• If you encounter MPS errors on Mac, try launching with: export PYTORCH_ENABLE_MPS_FALLBACK=1
• Or enable "Use CPU for SAM" checkbox in the UI (slower but more stable)

Memory Issues:

• Use smaller SAM models (sam_vit_b or mobile_sam) if you run out of VRAM
• Lower the detection threshold to get fewer detections

Contributing

If you encounter a bug, please submit an issue. Please include your WebUI version, extension version, browser version, and any relevant error logs.

Pull requests are welcome!

Credits

This is a simplified fork of sd-webui-segment-anything by continue-revolution.