[Coral_current] [IKDD News] 1Million Deepfakes Detection Challenge @ ACM Multimedia (fwd)

[Tim Oates]

An interesting competition.


---------------------------------------
Tim Oates, Professor
Department of CS and EE
University of Maryland Baltimore County
(410) 455-3082
https://coral-lab.umbc.edu/oates/

---------- Forwarded message ----------
Date: Thu, 7 Mar 2024 04:29:41 -0800 (PST)
From: Abhinav Dhall <dhallabhinav at gmail.com>
To: IKDD News <ikdd-news at googlegroups.com>
Subject: [IKDD News] 1Million Deepfakes Detection Challenge @ ACM Multimedia

Call for Participation - 1 Million Deepfakes Detection
Challengehttps://deepfakes1m.github.io/ 
at ACM Multimedia 2024, Melbourne

The tremendous progress in generative AI has made the generation and
manipulation of synthetic data easier and faster than before. To this end,
multiple use cases are benefitting from it. The negative aspect of this
progress and wide adoption of generative AI is deepfakes. Audio/image/video
of an individual(s) is manipulated using generative methods without
permission from the individual(s). This can make them be shown saying or
doing something, which they may not have done in real. These unethically
manipulated videos, popularly known as deepfakes have wide repercussions and
negative effects on society in the form of the deepfakes? potential in
spreading disinformation and misinformation. Deepfakes unfortunately are
used for trolling online as well. Authentication systems such as video KYC
(Know Your Customer) are also not resilient as often face recognition and
verification systems are deceived when high-quality deepfakes are used. To
this end, it is important for platforms and systems to be able to identify
if manipulation has been performed on a media. These systems, which detect
and analyse the deepfakes are referred to as deepfakes detectors.

The 1M-Deepfakes Detection Challenge comprises of two sub-tasks:

a. Deepfake Detection ? Given an audio-visual sample containing a single
subject, the task is to identify if the video is a deepfake or real.

b. Deepfakes Temporal Localisation ? Given an audio-visual sample containing a
single subject, the task is to find out the frames (time stamps) in which
the manipulation is done. The assumption here is that from the perspective
of spreading misinformation, editing a few vital parts of a video may be
enough to change the meaning of the original video, and at the same time,
the quality of the deepfake video will be closer to the original compared to
a deepfake in which the entire original video is manipulated.

Challenge Registration - https://deepfakes1m.github.io/

Timeline
Training and Validation Data - available now
Test Data - mid-May
Paper submission deadline - June 14

Thanks,
Abhinav Dhall

--
You received this message because you are subscribed to the Google Groups
"IKDD News" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to ikdd-news+unsubscribe at googlegroups.com.
To view this discussion on the web visithttps://groups.google.com/d/msgid/ikdd-news/6d12bbb1-5d65-4ad1-82fa-3c17c26
dec7an%40googlegroups.com.