This ComfyUI custom node provides advanced image background removal and segmentation capabilities, targeting users who need precise control over object, face, clothing, and fashion extraction. It integrates multiple state-of-the-art models, offering flexibility and high-quality results for complex image processing tasks within the ComfyUI ecosystem.

How It Works

The node leverages a modular design, allowing users to select from various segmentation models including RMBG-2.0, INSPYRENET, BEN, BEN2, BiRefNet variants, SAM, and GroundingDINO. It supports text-prompted segmentation via GroundingDINO and SAM, enabling users to specify target objects with natural language or tag-style prompts. The implementation focuses on providing fine-grained control over mask generation, including sensitivity, resolution, blur, and offset parameters, alongside options for background color and mask inversion.

Quick Start & Requirements

• Installation: Can be installed via ComfyUI-Manager, cloning the repository, or using comfy node install ComfyUI-RMBG.
• Prerequisites: ComfyUI, Python 3.10+. Automatically installs torch, torchvision, Pillow, numpy, huggingface-hub, tqdm, transformers, transparent-background, opencv-python. Models are downloaded automatically or can be manually placed in specified ComfyUI model directories.
• Links: update.md

Highlighted Details

• Supports high-resolution processing up to 2048x2048 with BiRefNet-HR.
• Includes specialized nodes for face, clothes, and fashion segmentation.
• Offers text-prompted object segmentation using SAM and GroundingDINO.
• Features internationalization (i18n) support for multiple languages.

Maintenance & Community

The project shows active development with frequent updates and new feature additions, including recent improvements to compatibility and new utility nodes. Links to community resources are not explicitly provided in the README.

Licensing & Compatibility

• License: GPL-3.0 License.
• Compatibility: The GPL-3.0 license is a strong copyleft license, which may impose restrictions on linking with closed-source software.

Limitations & Caveats

The README indicates that some updates require manual installation of requirements.txt. While models are generally auto-downloaded, manual download instructions are also provided, suggesting potential network or path issues for some users. The project is actively developed, implying potential for breaking changes between versions.