Tools
Tools
Deshpande, Saloni (2024) Hybrid Bayesian CNN-GAN architecture for Deepfake detection. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (849kB) | Preview |
Preview |
PDF (Configuration Manual)
Download (9MB) | Preview |
Abstract
Deepfakes have a huge impact on from multiple standpoints. Current deep fake detection tools lack real-time detection and accuracy to accommodate the classification speed between classifying the data into real and fake. Many researchers have tried to implement various models like hybrid CNN-LSTM or ResNet, MesoNet, or InceptionNet V2 but each model has its drawbacks when it comes to attention to detail thus we proposed a new hybrid model of CNN-GAN with a Bayesian approach for its architecture and compared it against two models. This work aims to contribute a new model and in doing so we have implemented our new model on the famous Deepfake Detection Challenge (DFDC) dataset by using face extraction from video frames and then our model was evaluated using a Confusion Matrix and comparing the accuracy results to most commonly used deep learning models. We have discovered that our proposed model has 80% accuracy and works within the desired speed of the pipeline. Thus, we conclude our experiment on deep fake detection using deep learning.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Jain, Mayank UNSPECIFIED |
| Uncontrolled Keywords: | Deep fake; CNN; Generative Adversarial Networks; autoencoder; deep learning; Bayesian; detection |
| Subjects: | Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning |
| Divisions: | School of Computing > Master of Science in Artificial Intelligence |
| Depositing User: | Ciara O'Brien |
| Date Deposited: | 18 Jun 2025 11:26 |
| Last Modified: | 18 Jun 2025 11:26 |
| URI: | https://norma.ncirl.ie/id/eprint/7908 |
Actions (login required)
 |
View Item |