ComfyUI-Watermark-Detection-YOLO

ComfyUI Watermark Detection Custom Node

Advanced watermark detection using a custom trained YOLO11 model with special support for fading watermarks.

🚀 Features

• High Accuracy: Custom model trained on 24,558 watermark images. Thanks to https://universe.roboflow.com/idevelop/watermarks-fczjj/dataset/1
• Multi-Resolution: Supports 640px, 1280px, 1920px detection
• Fading Watermarks: Low confidence threshold for transparent watermarks
• Binary Masks: Clean mask output for post-processing
• Detailed Info: Bounding box coordinates and confidence scores

📦 Installation

Via ComfyUI Manager (Recommended)

1. Open ComfyUI Manager
2. Search for "Watermark Detection Custom"
3. Click Install

Manual Installation

cd ComfyUI/custom_nodes/
git clone https://github.com/soliton80/ComfyUI-Watermark-Detection-YOLO.git
cd ComfyUI-Watermark-Detection-YOLO
pip install -r requirements.txt

🎯 Usage

1. Load Watermark Detector: Uses the trained watermarks_detect_best model
2. Detect Watermarks: Process images with configurable confidence/IOU
3. Save Results: Export detection images and masks
4. Confidence pointer: Hold lower confidence pointer for weak visible watermarks
5. Resolution Option: Put resolution higher for weak visible watermarks

📊 Model Performance

Trained on diverse watermark dataset including:

• Text overlays and logos
• Semi-transparent watermarks
• Various positions and sizes
• Multi-language text watermarks

📥 Example workflow

🤝 Issues

Issues are welcome!

📄 License

MIT License - see LICENSE file for details.