Sai, G Mohan and Noel, B Olive and Kiran, B L Sai and Krishna, B Karthik and Bhupal, M Ram (2024) Digital Image Authenticity Assessment Using Deep Learning. Journal of Engineering Research and Reports, 26 (5). pp. 283-293. ISSN 2582-2926
Sai2652024JERR115718.pdf - Published Version
Download (327kB)
Abstract
Its super important to find fake pictures online to make sure things are true so we came up with a smart idea using something called a convolutional neural network, CNN is like a really smart detective for pictures. It checks out lots of pictures, some real and some fake and learns how to tell them apart by looking at all the tiny details Before the detective work, we make all the pictures the same size, like putting them in the same-sized frame We also use a special tool called error level analysis (Ela). Ela helps us by showing parts of the pictures that might have been changed its like searching for clues in a detective movie We looked at a bunch of pictures, over 12,000 of them We made sure to have a mix of real ones like trees and people 7492 of them and some that were changed to trick people, 5123 of them this way, our detective could learn from lots of different situations our system is really good at finding fake pictures because it uses fancy math and clever techniques We tested it a lot to make sure it works in different situations, like when someone copies part of a picture or changes how it looks and guess what? Our tests showed that our system is great at finding fakes In the end, our idea helps make sure pictures online are real, which is super important in today’s world of digital pictures.
Item Type: | Article |
---|---|
Subjects: | Institute Archives > Engineering |
Depositing User: | Managing Editor |
Date Deposited: | 27 Apr 2024 09:49 |
Last Modified: | 27 Apr 2024 09:49 |
URI: | http://eprint.subtopublish.com/id/eprint/4254 |