Reproduce “SynthForensics: Benchmarking and Evaluating People-Centric Synthetic Video Deepfakes” 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
Roberto Leotta, Salvatore Alfio Sambataro, Claudio Vittorio Ragaglia, Mirko Casu, Yuri Petralia, Francesco Guarnera, Luca Guarnera, Sebastiano Battiato
SynthForensics introduces a people-centric benchmark for evaluating detectors on synthetic videos. It pairs generated clips with real-source videos, using validated prompts and outputs from eight text-to-video and seven image-to-video open-source generators. In paired ratings, participants preferred its videos to benchmark videos for quality in 71.5% and realism in 76.5% of comparisons. FTCN scored 99.70% AUC on FF++ but 44.76% on paired SF-FF++, showing limited transfer in this benchmark setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.