Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 10

INTERNATIONAL CONFERENCE

on
International Conference on Smart Innovations for Society
ICSIS-2022

Organized by:
Department of Applied Science, Poornima Institute
of Engineering and Technology, Jaipur

Unsupervised Machine Learning Algorithm


for Face Morphing

Presented by:
Hitesh Kumar Sharma

Authors:,
Hitesh Kumar Sharma, Mahadev, Raman Chadha,
Gaurav Nagarkoti, Prashant Ahlawat
MORPHING
• Face morphing is the process of morphing or merging
the face of subject image with the face of another test
image.

• Techniques of face morphing have been utilized in


quite diverse fields, mostly in the entertainment
industry to replace the face of an actor with a
resembling actor to complete movies in absence of
the actor.
RESEARCH AIM
• Our research work aims to design and implement a similar
face morphing program that has a model integrated into it
for selecting the best suitable candidate for face morphing.

• Using python programming and concepts of unsupervised


learning we aim to create will have the ability to calculate
the resemblance between the faces of the subject and
various candidate and decide for itself which among
different candidates is best suitable for face morphing.
PROPOSED SYSTEM
IMPLEMENTATION and RESULTS
Face Resemblance:

• Step 1 : Image is loaded and read via python programming

• Step 2 : Face is detected in the image using the OPENCV package.

• Step 3 : After Face detection , the face points and Face Encodings are of an Image is taken
into separate arrays.

• Step 4 : The euclidean distance between the face point is calculated for both source and
subject image and taking into account the face encoding, resemblance between two face is
found out in terms of true or false via FACE COMPARE feature of the Face_Recognition
Module.

• Step 5 : Program is developed to compare the images of multiple candidates with subject
image to find out the candidate with highest resemblance with the subject.
Face Morphing:

• Step 1: Face Points are detected for both the source and subject image

• Step 2 : On the basis of face points triangle mesh via Delaunay Triangulation
is formed on the faces of source image and the subject image.

• Step 3: The triangles of the source image is fit into the triangulations of the
subject image and the face in source image is reconstructed on the subject
image thus performing Morphing process.

• Step 4 : After the source image’s face has been reconstructed on the
subjects face , the color needs to adjusted to attain seamless results .
CONCLUSION
• Our research work was successful to build a program that
selects candidate for the face morphing procedure by
running the data of the candidate through 3 layers of test
where these candidates are filtered out and is successful in
producing the result as the candidate best for the
morphing procedure along with the best morphed output

• Hence resulting in a smart system that on it own selects


the best candidate for face morphing and also provides a
provisional morphed image as result.
THANK YOU
Queries……

You might also like