Get the latest tech news
Watermark Anything
Official implementation of the paper "Watermark Anything with Localized Messages" - facebookresearch/watermark-anything
For training our models we use the COCO dataset, with additional safety filters and where faces are blurred. Increasing it will lead to worse images but more robust watermarks, and vice versa. By default, it is set to 2.0, feel free to increase or decrease it to test how it influences the metrics.
Or read this on Hacker News