3D data augmentation and dual-branch model for robust face forgery detection

Zhou, Changshuang; Li, Frederick W.B.; Song, Chao; Zheng, Dong; Yang, Bailin

Abstract

We propose Dual-Branch Network (DBNet), a novel deepfake detection framework that addresses key limitations of existing works by jointly modeling 3D-temporal and fine-grained texture representations. Specifically, we aim to investigate how to (1) capture dynamic properties and spatial details in a unified model and (2) identify subtle inconsistencies beyond localized artifacts through temporally consistent modeling. To this end, DBNet extracts 3D landmarks from videos to construct temporal sequences for an RNN branch, while a Vision Transformer analyzes local patches. A Temporal Consistency-aware Loss is introduced to explicitly supervise the RNN. Additionally, a 3D generative model augments training data. Extensive experiments demonstrate our method achieves state-of-the-art performance on benchmarks, and ablation studies validate its effectiveness in generalizing to unseen data under various manipulations and compression.

Citation

Zhou, C., Li, F. W., Song, C., Zheng, D., & Yang, B. (2025). 3D data augmentation and dual-branch model for robust face forgery detection. Graphical Models, 138, Article 101255. https://doi.org/10.1016/j.gmod.2025.101255

• Journal Article Type: Article
• Acceptance Date: Jan 14, 2025
• Online Publication Date: Feb 4, 2025
• Publication Date: 2025-03
• Deposit Date: Feb 12, 2025
• Publicly Available Date: Feb 12, 2025
• Journal: Graphical Models
• Print ISSN: 1524-0703
• Electronic ISSN: 1524-0711
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 138
• Article Number: 101255
• DOI: https://doi.org/10.1016/j.gmod.2025.101255
• Public URL: https://durham-repository.worktribe.com/output/3474601