HumanSAM: Classifying Human-centric Forgery Videos in Human Spatial, Appearance, and Motion Anomaly

TL;DR

HumanSAM is a novel framework that classifies human-centric forgery videos into spatial, appearance, and motion anomalies, improving interpretability and robustness in forgery detection.

Contribution

The paper introduces HumanSAM, the first benchmark dataset for human-centric forgery videos, and proposes a new classification framework with a fusion of video understanding and depth features.

Findings

HumanSAM outperforms state-of-the-art methods in classification accuracy.

The HFV dataset provides comprehensive annotations for forgery types.

The rank-based confidence strategy enhances model robustness.

Abstract

Numerous synthesized videos from generative models, especially human-centric ones that simulate realistic human actions, pose significant threats to human information security and authenticity. While progress has been made in binary forgery video detection, the lack of fine-grained understanding of forgery types raises concerns regarding both reliability and interpretability, which are critical for real-world applications. To address this limitation, we propose HumanSAM, a new framework that builds upon the fundamental challenges of video generation models. Specifically, HumanSAM aims to classify human-centric forgeries into three distinct types of artifacts commonly observed in generated content: spatial, appearance, and motion anomaly. To better capture the features of geometry, semantics and spatiotemporal consistency, we propose to generate the human forgery representation by fusing two branches of video understanding and spatial depth. We also adopt a rank-based confidence enhancement strategy during the training process to learn more robust representation by introducing three prior scores. For training and evaluation, we construct the first public benchmark, the Human-centric Forgery Video (HFV) dataset, with all types of forgeries carefully annotated semi-automatically. In our experiments, HumanSAM yields promising results in comparison with state-of-the-art methods, both in binary and multi-class forgery classification.