iSchool Capstone

Pixel Patrol

Project tags:

The proliferation of AI-generated images and deepfakes on social media platforms has fueled misinformation and ethical concerns. This project addressed the need for reliable detection tools by developing an AI Image Detector Chrome extension to inform users if celebrity photos on Instagram are likely AI-generated. Transfer learning techniques were employed to fine-tune pre-trained CNN models on a dataset of real and AI-generated celebrity images. While performance analysis revealed inconsistencies, progress was made in generating AI images, training models, and developing front-end functionality. Moving forward, we aim to expand the dataset, incorporate explainable AI features, enable mobile integration, and conduct user-centric refinement. Our work strives to provide a transparent tool for identifying AI-generated content, empowering users to combat misinformation, and promoting ethical online practices.

Project participants:

Vickie Kuan Knight

Informatics

Daisy Shiewoo Lee

Informatics

Daphni Ann George

Informatics

Chantria Im

Informatics