RAW: Robust Avatar Watermarking -- Benchmarking and Baseline
Researchers have introduced RAW, a new benchmark designed to evaluate the robustness of digital avatar watermarking techniques. Existing methods struggle with common avatar post-processing steps like background replacement, which significantly degrade watermark recovery. To address this, the team also proposed WALT, a novel watermarking approach that embeds information in UV texture space through 3D face reconstruction, demonstrating superior performance against various attacks. AI
IMPACT This research could improve the security and provenance of digital avatars, crucial for the growing metaverse and AI-generated content.