Scribd
Upload
17 views2 pages

Project Abstract

Uploaded by

SaMPaTH CM 19&[
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
17 views2 pages

Project Abstract

Uploaded by

SaMPaTH CM 19&[
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Department of Computer Science and Engineering

PROJECT ABSTRACT –[ADMITTED BATCH: 2020-2024]


YEAR & BRANCH: BATCH NO.: A1 SECTION: CSE-1
IV B.TECH – COMPUTER SCIENCE AND ENGINEERING
BATCH DETAILS:
S. No. Student Regd. No. Student Name
1 20JG1A0525 G.Amrutha
2 20JG1A0543 K.Srija
3 20JG1A0506 A.Pujitha
4 20JG1A0528 K.Harshita Meghana

PROJECTTITLE:
PROJECT GUIDE NAME: “Recent Generative Adversarial Approach in Face Aging”
Dr. P.V.S.L. Jagadamba

ABSTRACT:
This paper focuses on the challenges faced by existing conditional Generative Adversarial Networks (GANs) in
accurately capturing the transition patterns between adjacent age groups during face aging. The authors propose a
novel approach called Contextual Generative Adversarial Nets (C-GANs), which includes a conditional
transformation network and two discriminative networks. The conditional transformation network mimics the aging
process using specially designed residual blocks, while the age discriminative network ensures alignment with real
conditional distribution. A unique transition pattern discriminative network is introduced to distinguish real
transition patterns from fake ones, acting as an additional regularization term. Experimental results showcase the
effectiveness of the proposed framework in producing compelling results compared to state-of-the-art methods and
ground truth, with observed performance gains in cross-age face verification.

DOMAINS WHERE THE PROJECT CAN BE IMPLEMENTED:


Forensic Science
Entertainment and Digital Arts
Healthcare
Biometrics and Security
Retail and Fashion

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING Page 1


IMPLEMENTATION:
1. Develop an architecture tailored for face aging. Design a conditional transformation network, incorporating
specially designed residual blocks to simulate the aging process.
Algorithm: Contextual Generative Adversarial Network (C-GAN)/Cyclic GAN.
2. Train the GAN on the prepared dataset using the designed architecture. Implement loss functions that
consider both the age discrimination and the realism of transition patterns.
Technique: Keras Functions/PyTorch Functions.
3. Experiment with different hyperparameter configurations to optimize the performance of the GAN. Adjust
learning rates, batch sizes, and regularization parameters.
Regularization Techniques: L1 or L2 Regularizers.
4. Evaluate the trained model on a separate test dataset and compare the generated aged faces with ground truth
images.
Evaluation Metrics: MSE, Confusion Matrix.
5. Create a user-friendly interface for users with system and display generated faces based on the Specific age
groups.

PRINCIPLE IDEA:
The fundamental idea behind face aging using GANs is to leverage the power of Generative Adversarial Networks
to simulate and generate realistic facial aging transformations. By training the GAN on a dataset containing pairs of
facial images at different ages, the model learns complex relationships between young and aged faces. The
conditional transformation network within the GAN architecture imitates the aging process, incorporating specially
designed blocks to capture nuanced changes. The age discriminative network guides the synthesis to align with real
age distributions, while a novel transition pattern discriminative network ensures the model captures the gradual
shape and texture changes between adjacent age groups. This principle idea aims to produce contextually accurate
and visually compelling age-progressed faces.

VISION:
The goal of implementing a face aging project using GANs is to accurately predict and generate realistic aged facial
images from input young faces, leveraging contextual information like facial landmarks or attributes. This process
aims to create a versatile tool for various applications, including digital entertainment, forensics, and personalized
cosmetics, enhancing user experiences and facilitating tasks like age progression in photographs or facial recognition
across different age groups.

BASE REFERENCE PAPER DETAILS:


TITLE: Recent Generative Adversarial Approach in Face Aging and Dataset Review
PUBLISHED YEAR: 2022
JOURNAL NAME: IEEE Access

Signature of Head of Department Name &Signature of Project Guide

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING Page 2

You might also like