Deepfake Detection AI

ResNet18-based synthetic media classifier

SYSTEM OVERVIEW

Deep learning model using ResNet18 fine-tuned on the FaceForensics++ dataset. Achieves 94% accuracy on detecting AI-generated face videos. Includes a web interface for live classification.

ARCHITECTURE

INPUT → PREPROCESSING → MODEL

↓

INFERENCE → POST-PROCESSING

↓

OUTPUT → VISUALIZATION → API

TECH STACK

PythonPyTorchResNet18OpenCVCUDAFlask

LINKS

GitHub — Private