Reproduce “MARS: Multiscale Appearance-Residual Streams for Video Forgery Detection” 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
Ahmed Y. Radwan, Shaina Raza
MARS detects video forgeries involving face edits and fully generated scenes. It combines appearance features with multiscale spatial and temporal differences in frozen features, alongside an adapted semantic path. Under shared FF++ plus GTA-V training, it reaches 91.43% video-level AUROC across seven datasets, versus 90.03% for DFD-HR. Results vary across forgery categories.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.