ComicContours: Facial Feature Augmentation Using GANs

CS 426 Senior Projects in Computer Science, Spring 2024

University of Nevada, Reno, Department of Computer Science and Engineering

Project Overview

ComicContours represents the culmination of efforts by Team 7 in the CS 426 Senior Projects course at the University of Nevada, Reno. This project introduces an innovative approach to digital art creation, employing Generative Adversarial Networks (GANs) to transform user-uploaded images into caricatures. Our approach not only facilitates a novel form of artistic creativity but also contributes to the fields of computer vision and machine learning.

Team Members

• Rhiannon Barber
• Karly Amanna
• JoJo Petersky

Faculty and External Advisors

• Instructors: Sara Davis, Vinh Le, Dave Feil-Seifer
• External Advisors: Dr. Emily Hand, Sara Davis

Project Description

ComicContours aims to democratize the process of caricature creation through the application of advanced machine learning techniques, specifically GANs. By automating the traditionally manual and skill-intensive process of caricature drawing, our project opens up new avenues for artistic expression and digital avatar customization. We target a diverse audience ranging from digital art enthusiasts and social media users to professional graphic designers seeking innovative representation methods.

Our technology stack includes PyCharm IDE and libraries such as NumPy, DLib, OpenCV for facial recognition and augmentation, alongside tkinter, shutil, and PIL for the user interface. Security measures are upheld using bcrypt, sqlite3, and email-related libraries to ensure user data protection.

We are committed to regular updates and you can stay informed on our progress, including updates on project deliverables, resources, and news, by navigating through our website.

Related Resources

Our project draws inspiration and knowledge from a range of sources within the domain of digital art generation and machine learning. Below are some key resources that have informed our work:

• Problem Domain Book:
  • “Handbook of Face Recognition” by Stan Z. Li and Anil K. Jain. A comprehensive guide exploring the methods and challenges in the field of face recognition. View on Springer
• Useful Websites:
  • Face++: An advanced facial recognition platform that offers deep learning algorithms for feature enhancement and identification. Visit Face++
  • DeepArt.io: Utilizes an innovative AI model to transform images into artwork mimicking famous styles. Explore DeepArt
• Technical Reports, Conference Papers, Journal Articles:
  • “DeepFace: Closing the Gap to Human-Level Performance in Face Verification” by Yaniv Taigman et al. Discusses the DeepFace model’s approach to improving face verification performance. Read on IEEE Xplore
  • “VGGFace2: A dataset for recognising faces across pose and age” by Q. Cao et al. Introduces a diverse dataset aimed at improving face recognition technology. Access on IEEE Xplore
• News and Other Information:
  • MIT Technology Review’s article on hyper-realistic beauty filters explores the impact and technology behind AI-generated filters in social media. MIT Technology Review Article

For more detailed information on the technologies and methodologies we’re exploring in our project, please refer to the links provided.

Legal and Ethical Considerations

In adherence to legal and ethical standards, ComicContours incorporates a user consent mechanism for photo uploads and emphasizes the security and privacy of user data. We prioritize privacy and security, incorporating consent mechanisms and adhering to the ACM Code of Ethics. Our commitment extends to ADA compliance, ensuring ComicContours is accessible, maintaining transparency, and fostering an inclusive digital environment.

Updates and Progress

As ComicContours evolves, we commit to providing regular updates on our website. Recent developments include enhancements to the user interface and security features, along with progress in facial feature augmentation and blending techniques.

Accessibility Commitment

ComicContours is designed with accessibility in mind, following WCAG guidelines to ensure inclusivity for all users. We are actively working to implement features like text resizing and keyboard navigability to accommodate diverse needs.

This website is committed to being available for at least one year post-semester, providing a lasting resource for future CS 426 students and interested parties.

For more information or to get in touch, please contact us at [email protected].