๐ MimiLens Watermark Detector
Detect watermarks, signatures, speech bubbles, text, and manga effects in images using the YOLO-based detector from Minthy/MimiLens-Models.
Upload an image to find persistent artifacts that may contaminate generative model training data.
Examples
About: The detector uses YOLOv26 (detect_y26l_v1.2.pt) trained to recognize 5 classes:
watermarks, signatures, regular text, speech bubbles, and comic/manga effects.
Post-processing merges nearby watermark/signature detections and removes text inside speech bubbles.
The YOLO model is AGPL-3.0 licensed; see the model card for details.