Reproduce “Stacking Brick by Brick: Aligned Feature Isolation for Incremental Face 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
Jikang Cheng, Zhiyuan Yan, Ying Zhang, Li Hao, Jiaxin Ai, Qin Zou, Chen Li, Zhongyuan Wang
SUR-LID updates face-forgery detectors as new forgery datasets arrive, aiming to preserve earlier patterns while learning new ones. Sparse Uniform Replay retains a spread of old examples, while the Latent-space Incremental Detector separates old and new features and aligns real/fake boundaries. On Protocol 2, its final mean frame-level AUC is 0.9433 versus DFIL’s 0.7929, though DFIL scores higher on EFS.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.