AVT$^{2}$-DWF: Improving Deepfake Detection With Audio-Visual Fusion and Dynamic Weighting Strategies
Rui Wang, Dengpan Ye, Long Tang, Yunming Zhang, Jiacheng Deng
Abstract
With the continuous improvements of deepfake methods, forgery messages have transitioned from single-modality to multi-modal fusion, posing new challenges for existing forgery detection algorithms. In this letter, we propose <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">AVT<inline-formula><tex-math notation="LaTeX">$^{2}$</tex-math></inline-formula>-DWF</b>, the <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A</b>udio-<bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</b>isual dual <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">T</b>ransformers grounded in <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D</b>ynamic <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W</b>eight <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F</b>usion, which aims to amplify both intra- and cross-modal forgery cues, thereby enhancing detection capabilities. AVT<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{2}$</tex-math></inline-formula>-DWF adopts a dual-stage approach to capture both spatial characteristics and temporal dynamics of facial expressions. This is achieved through a face transformer with an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$n$</tex-math></inline-formula>-frame-wise tokenization strategy encoder and an audio transformer encoder. Subsequently, it uses multi-modal conversion with dynamic weight fusion to address the challenge of heterogeneous information fusion between audio and visual modalities. Experiments on DeepfakeTIMIT, FakeAVCeleb, and DFDC datasets indicate that AVT<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{2}$</tex-math></inline-formula>-DWF achieves state-of-the-art performance intra- and cross-dataset Deepfake detection.