ComfyUI-GROUT

ComfyUI-GROUT

Custom ComfyUI nodes for GROUT (Geometric Reasoning Over Unstructured Tessellations) - a deep learning model for detecting and segmenting grout lines in mosaic images.

Nodes

GROUT Segmentation

Main segmentation node that processes mosaic images.

Inputs:

• image: Input mosaic image
• device: Select CUDA or CPU for inference
• threshold: Initial threshold for binary mask (0.0 - 1.0)
• model_path (optional): Custom path to model weights. Leave empty to auto-download from HuggingFace.

Outputs:

• probability_heatmap: Grayscale probability map (0-1 range)
• binary_mask: Thresholded binary mask as image
• mask: ComfyUI MASK format for compositing

GROUT Threshold

Post-processing node to adjust threshold without re-running inference.

Inputs:

• probability_heatmap: Output from GROUT Segmentation node
• threshold: Threshold value (0.0 - 1.0)

Outputs:

• binary_mask: Thresholded binary mask as image
• mask: ComfyUI MASK format

GROUT Heatmap Colorize

Visualization node to apply colormap to probability heatmap.

Inputs:

• probability_heatmap: Output from GROUT Segmentation node
• colormap: Select from magma, viridis, plasma, inferno, hot, jet

Outputs:

• colored_heatmap: Colorized visualization

Installation

Option 1: ComfyUI Manager

Search for "GROUT" in ComfyUI Manager and install.

Option 2: Manual Installation

1. Navigate to your ComfyUI custom nodes folder:

   cd ComfyUI/custom_nodes/
   

2. Clone this repository:

   git clone https://github.com/advancedtech-sk/ComfyUI-GROUT.git
   

3. Install dependencies:

   cd ComfyUI-GROUT
   pip install -r requirements.txt
   

   For standalone Windows version (portable/embedded Python):

   cd ComfyUI-GROUT
   ..\..\..\python_embeded\python.exe -m pip install -r requirements.txt
   

4. Restart ComfyUI

Model Download

The model will be automatically downloaded from HuggingFace on first use. It will be saved to:

ComfyUI/models/grout/grout_b3_zeroshot_v1.pth

Alternatively, you can manually download from:

• HuggingFace Model

Example Workflow

[Load Image] -> [GROUT Segmentation] -> [probability_heatmap] -> [GROUT Threshold] -> [binary_mask]
                         |
                         +-> [GROUT Heatmap Colorize] -> [Preview Image]

Workflow Tips:

1. Use GROUT Segmentation to run inference once
2. Connect the probability_heatmap output to GROUT Threshold
3. Adjust the threshold slider in GROUT Threshold to tune the mask without re-running the model
4. Use GROUT Heatmap Colorize for visualization with different colormaps

Links

• GROUT Paper (Zenodo)
• GROUT Model (HuggingFace)
• GROUT Demo (HuggingFace Spaces)
• GROUT GitHub

License

MIT License

Author

Radoslav Lovecky, Arperon s.r.o, Institute of Advanced Technologies, Slovakia