This project provides extensions for Automatic1111 and ComfyUI to automatically generate segmentation masks for various human body parts (background, hair, body, face, clothes) within an image-to-image workflow. It leverages Google's Multi-class selfie segmentation model, offering a streamlined way to isolate and manipulate specific elements of a person in generated images.

How It Works

The extension utilizes Google's Multi-class selfie segmentation model to perform pixel-level classification of human subjects. It generates distinct masks for background, hair, body, face, and clothes. A recent refinement feature adds a bounding box around detected masks, crops the image to that area, and re-runs segmentation for improved accuracy, particularly on distant subjects or high-resolution images.

Quick Start & Requirements

• ComfyUI: git clone https://github.com/djbielejeski/a-person-mask-generator into custom_nodes folder, then pip install -r requirements.txt.
• Automatic1111: Install via the "Install from URL" tab in the Extensions menu with https://github.com/djbielejeski/a-person-mask-generator.git.
• Prerequisites: Python, Git. ComfyUI or Automatic1111 Web UI.
• Resources: Requires sufficient VRAM for the segmentation model.

Highlighted Details

• Supports both ComfyUI and Automatic1111 Stable Diffusion Web UI.
• Generates masks for background, hair, body, face, and clothes.
• Includes a mask refinement option for improved accuracy.
• Offers facial landmark mask outputs for ComfyUI.

Maintenance & Community

The project has seen recent updates, including ComfyUI compatibility fixes and mask refinement features. There are no explicit links to community channels or roadmaps provided in the README.

Licensing & Compatibility

The repository does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The project does not specify a license, which may impact commercial use. There are no explicit benchmarks or performance metrics provided.