BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//RLASKEY//CALENDEROUS//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20261017T164338Z
LAST-MODIFIED:20200814T180021Z
DTSTART:20200807T163000Z
DTEND:20200807T173000Z
UID:event2351@bu.edu
URL:http://physics.bu.edu/internal/events/show/2351
SUMMARY:Using CNN to Detect Fake News
DESCRIPTION:Featuring Julian Ingham\, Boston University\nHosted by: Eric Bo
	yers\n\nPart of the Graduate Student Council Events.\n\nPart of the student
	 seminar series. A recording will be posted on the event page.\n\nAbstract:
	 Recent advances in machine learning techniques have made possible the crea
	tion of high quality falsified videos. One particular type of fake videos\,
	 commonly known as DeepFakes\, has recently drawn especial attention. In a 
	DeepFake\, the faces of a target individual are replaced by the faces of a 
	donor individual synthesized by a neural network\, retaining the target’s
	 facial expressions and head poses. Well-crafted DeepFakes can create highl
	y convincing illusions of a person’s presence and activities -- threateni
	ng serious political\, social\, and legal consequences. This talk will repo
	rt on some investigations into the task of discriminating between real vide
	o data and Deepfakes. A large volume of video data were processed into a se
	t of still color images. A type of neural network known as a Convolutional 
	Neural Network (CNN) was employed to detect artifacts of the forgery in the
	 facial region\, such as distortions and color blotches. Experimenting with
	 different network architectures\, we found the best results on our test se
	t were highly accurate\, among the top 100 in the Kaggle Deepfake Detection
	 Challenge (out of 2\,265 entrants). Further\, we found that certain kinds 
	of manipulation of the input data (rotation and dilatations) decrease predi
	ctive power\, suggesting the presence of asymmetric correlations in Deepfak
	e artifacts. We summarize preliminary additional experimentation with trans
	fer learning.
LOCATION: \, \, 
STATUS:CONFIRMED
CLASS:PUBLIC
END:VEVENT
END:VCALENDAR