ActivityForensics: A Comprehensive Benchmark for Localizing Manipulated Activity in Videos
This benchmark addresses the critical problem of detecting manipulated human actions in videos, which is essential for media authenticity and public trust, and provides a foundation for future research in temporal forgery localization.
ActivityForensics introduces the first large-scale benchmark for localizing activity-level forgeries in videos, containing over 6K forged segments that are visually indistinguishable from authentic content. The proposed Temporal Artifact Diffuser (TADiff) baseline achieves strong performance across intra-domain, cross-domain, and open-world settings.
Temporal forgery localization aims to temporally identify manipulated segments in videos. Most existing benchmarks focus on appearance-level forgeries, such as face swapping and object removal. However, recent advances in video generation have driven the emergence of activity-level forgeries that modify human actions to distort event semantics, resulting in highly deceptive forgeries that critically undermine media authenticity and public trust. To overcome this issue, we introduce ActivityForensics, the first large-scale benchmark for localizing manipulated activity in videos. It contains over 6K forged video segments that are seamlessly blended into the video context, rendering high visual consistency that makes them almost indistinguishable from authentic content to the human eye. We further propose Temporal Artifact Diffuser (TADiff), a simple yet effective baseline that exposes artifact cues through a diffusion-based feature regularizer. Based on ActivityForensics, we introduce comprehensive evaluation protocols covering intra-domain, cross-domain, and open-world settings, and benchmark a wide range of state-of-the-art forgery localizers to facilitate future research. The dataset and code are available at https://activityforensics.github.io.