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.
- Detects watermarks in images using YOLOv11
- Returns both the detection visualization and a binary mask
- Adjustable confidence threshold for detections
- Place this folder in your
ComfyUI/custom_nodes/
directory - The model will be automatically downloaded from Hugging Face on first use
- Add a
Load Watermark Detector
node to load the model - Connect your image to the
Detect Watermarks
node - Adjust the confidence threshold as needed (default: 0.5)
- The node outputs:
- Detection visualization (IMAGE)
- Binary mask of detected watermarks (MASK)
- Model: YOLOv11 trained on watermark detection
- Original Author: fancyfeast (JoyCaption creator)
- Model Source: Hugging Face
- Re-hosted at Hugging Face
- The model will be downloaded to
ComfyUI/models/yolo/
automatically - Higher confidence thresholds will result in fewer but more certain detections