DeepFake_Seminar.pptx

“DEEPFAKES”
Seminar On
SANDESH DIPAK BAGESHWAR
DR. A.B. DESHMUKH
Submitted by
Under the guidance of
DEPARTMENT OF INFORMATION TECHNOLOGY
SIPNA COLLEGE OF ENGINEERING AND TECHNOLOGY, AMRAVATI

What is Deepfake ?
2
Deepfake is synthetic media in which a person in an existing image
or video is replaced by someone's else likeness.

Contents
3
1 3 5
6
4
2
ORIGIN WORKING APPLICATION
DEVELOPMENT DETECTION CONCLUSION

1. ORIGIN
◎ This technology is invented in 2014 by Ian Goodfellow.
◎ The term deepfake originated around 2017 from a reddit user named
deep fakes.
◎ He as well as others in the reddit community(deepfakes) shared
deepfakes they created. There were many videos involved celebrity’s
faces swapped on to the body of actresses in pornographic videos.
◎ while non pornographic content included many videos with actor
nicolos cage’s face swapped into the various movies.
4

5
Amy Adams in the film Man of Steel and the deepfake created by YouTube
channel derpfakes that imposes actor Nicolas Cage's face over Adams'.

2. DEVELOPMENT
◎ In January 2018, a proper desktop application called “Fake App”
was launched. This app allows users to easily create and share videos
with their faces swapped with each other.
◎ In 2019 the mobile app giant Momo created the application Zao
which allows users to superimpose their face on television and movie
clips with a single picture.
◎ The Japanese Al company DataGrid made a full body deepfake that
can create a person from scratch.
6

2. DEVELOPMENT
◎ A mobile deepfake app “impressions” was launched in march 2020.
It was the first app for the creation of celebrity deep fake videos from
mobile phones.
◎ After these , large number of companies started to use deepfakes.
7

3. WORKING
◎ The program is fed a vast amount of data which it then uses to learn
to create its own ,new data.
◎ primarily they are based on autoencoders and Generative
adversarial networks (GAN).
8

3. 1 Autoencoders
◎ An autoencoder consists of 3 components : an encoder ,a code and
a decoder
◎ Encoder compresses the input data and produces the code after the
decoder reconstructs the input based only on the code.
◎ There are various types of Encoders : denoising encoders , deep
encoders, contractive encoders ,etc.
9

Autoencoders training phase
11

Autoencoders generation phase
12

3. 2 Generative Aadversarial Networks (GANs)
◎ GAN is an approach to generative modeling from input
dataset.These learn from data in order to generate new data.
◎ The system is trained by two distinctive neural networks : a
generator and a discriminator.
◎ The generator discovers regularities or patterns in the input dataset
and learns to reproduce them.The generated data set is sent to
discriminator coupled with real data for evaluation.
13

3. 2 Generative Aadversarial Networks (GANs)
14

4. Deepfake Detection
◎ Unnatural facial expressions.
◎ Unnatural eye movement.
◎ Awkward facial-feature positioning.
◎ A lack of emotion.
◎ Awkward-looking body or posture.
15

4. Deepfake Detection
◎ Unnatural body movement.
◎ Unnatural coloring.
◎ Hair that doesn't look real:
◎ Teeth that don't look real:
◎ Blurring or misalignment:
16

5. Positive Applications Of Deepfakes
◎ Education
◎ Art
◎ Autonomy & Expression
◎ Amplification of the Message and its Reach
◎ Innovation
17

6. Negative Applications Of Deepfakes
◎ Corporate Level Fraud
◎ Extorting Money from Businesses or Individuals
◎ False Information/Fake News
◎ Fake Videos
◎ Pornography
◎ Sockpuppets
◎ Blackmail
18

7. CONCLUSION
◎ Deep fake is very young and promising technology.
◎ Humanity is still getting acquainted with it and has not yet found its full
application within our society.
◎ Like many technologies it has its advantages and disadvantages.It could
harm or improve our world.
◎ We will need time to understand how to make the most out of it.
19

Thank You !
Don't be a viewer, be an analytic observer.
20

DeepFake_Seminar.pptx

More Related Content

What's hot (20)

Similar to DeepFake_Seminar.pptx (20)

Recently uploaded (20)

DeepFake_Seminar.pptx