The MeVer DeepFake Detection Service: Lessons Learnt from Developing and Deploying in the Wild

Baxevanakis, Spiros; Kordopatis-Zilos, Giorgos; Galopoulos, Panagiotis; Apostolidis, Lazaros; Levacher, Killian; Schlicht, Ipek B.; Teyssou, Denis; Kompatsiaris, Ioannis; Papadopoulos, Symeon

doi:10.1145/3512732.3533587

Cited by 5 publications

(3 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To tackle the challenge of deepfake detection in videos, many video-based deepfake detectors have been developed. Even if some approaches propose solutions which also analyse the temporal information of manipulated videos [8][9][10][11], the majority of methods are frame-based, classifying each video frame individually. Furthermore, several competitions have been organized to stimulate the resolution of this task including [12,13].…”

Section: Deepfake Detectionmentioning

confidence: 99%

“…The extracted frames are pre-processed, similar to many other deepfake detection methods [8][9][10]22,25] by introducing a face extraction step using the state-of-the-art face detector, MTCNN [37]. The models are trained and evaluated on a per-face basis and data augmentation was performed, similar to [8,21,25].…”

Section: The Followed Approach and The Tested Network Architecturesmentioning

confidence: 99%

See 1 more Smart Citation

On the Generalization of Deep Learning Models in Video Deepfake Detection

et al. 2023

View full text Add to dashboard Cite

The increasing use of deep learning techniques to manipulate images and videos, commonly referred to as “deepfakes”, is making it more challenging to differentiate between real and fake content, while various deepfake detection systems have been developed, they often struggle to detect deepfakes in real-world situations. In particular, these methods are often unable to effectively distinguish images or videos when these are modified using novel techniques which have not been used in the training set. In this study, we carry out an analysis of different deep learning architectures in an attempt to understand which is more capable of better generalizing the concept of deepfake. According to our results, it appears that Convolutional Neural Networks (CNNs) seem to be more capable of storing specific anomalies and thus excel in cases of datasets with a limited number of elements and manipulation methodologies. The Vision Transformer, conversely, is more effective when trained with more varied datasets, achieving more outstanding generalization capabilities than the other methods analysed. Finally, the Swin Transformer appears to be a good alternative for using an attention-based method in a more limited data regime and performs very well in cross-dataset scenarios. All the analysed architectures seem to have a different way to look at deepfakes, but since in a real-world environment the generalization capability is essential, based on the experiments carried out, the attention-based architectures seem to provide superior performances.

show abstract

Section: Deepfake Detectionmentioning

confidence: 99%

Section: The Followed Approach and The Tested Network Architecturesmentioning

confidence: 99%

On the Generalization of Deep Learning Models in Video Deepfake Detection

et al. 2023

View full text Add to dashboard Cite

show abstract

“…MTCNN [7]. An additional 30% surface area to include a portion of the background has been added to each detected face, as done in the literature [3,1]. To make sure that the subsequent preprocessing steps are not polluted by false detection, the face detection threshold is set at a rather high value of 95%.…”

Section: Additional Preprocessing Detailsmentioning

confidence: 99%