← Back to Projects

Deepfake Detection System

A multimodal deep learning system for deepfake video detection combining teeth detection and lip–audio synchronization.

Python TensorFlow Keras Computer Vision

Overview

A multimodal deep learning system for detecting deepfake videos, developed during MSc research at Otto-Von-Guericke University. The system fuses visual and audio signals to identify manipulated content.

Key Features

  • Teeth detection model for identifying visual inconsistencies across video frames
  • Lip–audio synchronization module for cross-modal consistency checking
  • Multimodal fusion pipeline combining visual and audio analysis signals

Results

Achieved 93.26% detection accuracy on a benchmark deepfake dataset using the combined multimodal approach.