Projects
/
Deepfake Speech Detection
Deepfake Speech Detection
Sep 2023 - Apr 2024 · 7 mos

Deepfake Speech Detection

Detecting deepfakes and bringing down misinformation
AI
Python
Signal Processing
Tensorflow
Pytorch
Cuda
AWS
Full Stack
Git

• Deployed a web app on AWS for journalists and AI researchers to run and retrain deepfake detectors respectively

• Benchmarked and ensembled multiple state of the art models and improved the resultant accuracy than the current state of the art by 30% on Singapore Speech Corpus

• Used energy based estimator to create confidence score and enhance out-of-distribution detection performance

← 
Back to all projects