This project was carried out as part of the TechLabs "Digital Shaper Program (Summer Term 2024)” in Aachen, Germany.
We discuss an approach based on remote photoplethysmography (rPPG) for Blood Volume Pulse (BVP) signal extraction from video input and the use of relevant features from this signal to develop a model which can classify real and fake videos.
Concepts from recent research papers were used in the development of algorithm for this project.
Python code for different ways of BVP extraction are available in all the main folders in this repo. Program code related to feature extraction and model training is available in folders /timo and /karthik.
Thank you!