Optimizing deepfake detection using deep convolutional neural networks for real-time detection
No Thumbnail Available
Date
2023-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Don Mariano Marcos Memorial State University - South La Union Campus
Abstract
In a world where there is such an abundance of video and image materials that are consumed every day on social media, the proliferation of falsified media is causing an increasing alarm.
To detect deepfakes using deep convolutional networks, a quasi-experimental research design was adapted to determine the preprocessing techniques, convolutional neural networks, and model that best perform in detecting deepfakes. Datasets were gathered from publicly available resources, then these datasets were pre-processed and used in training and testing the proposed model. Finally, the trained model’s performance was evaluated in terms of accuracy, precision, recall, f1-score, and speed in detecting deepfake images. By configuring these methods and architecture, the researcher determined which setting would be best in detecting deepfake images.
It was found that in preprocessing datasets for deepfake detection, data normalization, percentage splitting, inclusion or exclusion of data augmentation improves the training progress while randomization is avoided. On the other hand, the use of base models with deep and great number of layers and connections improves performance in classifying deepfakes.
Description
Abstract Only
Keywords
Citation
Bacungan, E.O. (2023). Optimizing deepfake detection using deep convolutional neural networks for real-time detection. [Unpublished Master's Thesis]. Don Mariano Marcos Memorial State University – South La Union Campus, Agoo, La Union. Lakasa ti Sirib, DMMMSU Institutional Repository.
Collections
Total visits
| views | |
|---|---|
| Optimizing deepfake detection using deep convolutional neural networks for real-time detection | 0 |