Reproduce “Made to Measure: Designing Image Watermarks to Specification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mingzhe Li, Yuefeng Peng, Kejing Xia, Pranav Jeyakumar, Ruolan Leslie Famularo, Shiqing Ma
TAILOR selects and validates combinations of existing image watermarks for requests specifying attacks, false-positive rate, image quality, and latency limits. It characterizes watermark fragments, jointly tunes their strengths and order, then checks candidates on user images. Across 7,321 requests in five scenarios, TAILOR-F achieved 96.21% average satisfaction, but only 85.40% in watermark-removal scenario S4 under the evaluated conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.