DeepFind!
Supervisor : Rose Gomar
Team size: Minimum 3, Maximum 5
| CSE | SE | Comm | Biomed | EE | Aero | Special |
|---|---|---|---|---|---|---|
| Yes | Yes | No | No | No | No | No |
Description
As artificial intelligence has evolved, the capabilities of sound, video, and image auto-generation, malicious users may seek to use artificial intelligence to mimic real persona’s and possibly use an assumed identity for fraudulent purposes. This type of mimicry is colloquially known as “Deep Faking”. AI created deep fakes can do all sorts of things, such as replacing a person's face in a video, clone voices, manipulate objects in videos, and even generate an entire model of a living person with advanced features such as mimicking tier posture, movement, gestures, etc. The DeepFind project aims to detect AI-generated (deepfake) videos in real time by implementing a machine learning algorithm accelerated on a Field-Programmable Gate Array (FPGA). The system will notify users of potentially deepfaked content by displaying an on-screen overlay.
Prerequisites:
This project has been taken.