Watermarking Images in Self-Supervised Latent Spaces
This work addresses the need for robust image watermarking techniques, offering an incremental improvement by leveraging self-supervised learning for enhanced performance.
The paper tackles the problem of watermarking images by embedding marks and binary messages into self-supervised latent spaces, achieving robustness to transformations like rotations and JPEG compression. It significantly outperforms previous zero-bit methods and matches state-of-the-art multi-bit watermarking performance.
We revisit watermarking techniques based on pre-trained deep networks, in the light of self-supervised approaches. We present a way to embed both marks and binary messages into their latent spaces, leveraging data augmentation at marking time. Our method can operate at any resolution and creates watermarks robust to a broad range of transformations (rotations, crops, JPEG, contrast, etc). It significantly outperforms the previous zero-bit methods, and its performance on multi-bit watermarking is on par with state-of-the-art encoder-decoder architectures trained end-to-end for watermarking. The code is available at github.com/facebookresearch/ssl_watermarking