ComfyUI Watermark Detection Node

This custom node for ComfyUI provides watermark detection capabilities using a YOLO model trained by fancyfeast, the creator of JoyCaption. The model is originally hosted at Hugging Face Space.

Features

• Detects watermarks in images using YOLOv11
• Returns both the detection visualization and a binary mask
• Adjustable confidence threshold for detections

Installation

1. Place this folder in your ComfyUI/custom_nodes/ directory
2. The model will be automatically downloaded from Hugging Face on first use

Usage

1. Add a Load Watermark Detector node to load the model
2. Connect your image to the Detect Watermarks node
3. Adjust the confidence threshold as needed (default: 0.5)
4. The node outputs:
   • Detection visualization (IMAGE)
   • Binary mask of detected watermarks (MASK)

Model Details

• Model: YOLOv11 trained on watermark detection
• Original Author: fancyfeast (JoyCaption creator)
• Model Source: Hugging Face
• Re-hosted at Hugging Face

Notes

• The model will be downloaded to ComfyUI/models/yolo/ automatically
• Higher confidence thresholds will result in fewer but more certain detections