Hierarchical Learning and Dummy Triplet Loss for Efficient Deepfake Detection - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue ACM Transactions on Multimedia Computing, Communications and Applications Année : 2023

Hierarchical Learning and Dummy Triplet Loss for Efficient Deepfake Detection

Résumé

The advancement of generative models has made it easier to create highly realistic Deepfake videos. This accessibility has led to a surge in research on Deepfake detection to mitigate potential misuse. Typically, Deepfake detection models utilize binary backbones, even though the training dataset contains additional exploitable information, such as the Deepfake generation method employed for each video. However, recent findings suggest that inferring a binary class from a multi-class backbone yields superior performance compared to directly employing a binary backbone. Building upon this research, our paper introduces two novel methods to infer a binary class from a multi-class backbone. The first method, named ”root dummies,” leverages the Dummy Triplet (DmyT) loss, which employs fixed vectors (i.e., dummies) instead of mined positives and negatives in the triplet loss. By training the multi-class backbone with these dummies, we can easily infer a binary class during testing by adjusting the number of dummies (from 6 during training to 2 during inference). Through this approach, we achieve an accuracy improvement of 0.23% compared to the existing inference method, without requiring additional training. The second proposed method is transfer learning. It involves training a classifier, such as a Support Vector Machine (SVM), to predict binary classes based on the image embeddings generated by the multi-class backbone. While this method necessitates additional training, it further enhances the model’s performance, resulting in an accuracy increase of 1.79%. In summary, our proposed methods improve the accuracy of Deepfake detection by simply modifying the number of classes during training, making them suitable for integration into a variety of existing Deepfake training pipelines. Additionally, to foster reproducible research, we have made the source code of our solution publicly available at https://github.com/beuve/DmyT.

Dates et versions

hal-04234773 , version 1 (10-10-2023)

Identifiants

Citer

Nicolas Beuve, Wassim Hamidouche, Olivier Déforges. Hierarchical Learning and Dummy Triplet Loss for Efficient Deepfake Detection. ACM Transactions on Multimedia Computing, Communications and Applications, 2023, ⟨10.1145/3626101⟩. ⟨hal-04234773⟩
13 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More