Back to search

Institutional Repository of Vilnius University

A system for recognising the ai manipulation on photographs /

Abstract

dc:description

KOSTENKO, Oleg. (2025) A System For Recognising The AI Manipulation On Photographs. Bachelor’s Thesis. Kaunas: Vilnius University Kaunas Faculty. 74 pages. SUMMARY The “RevealAI” information system has been developed with a goal of offering a trustworthy and transparent means of assessing the authenticity of a digital image and determining if it has been produced using artificial intelligence. In a world of increasing use of AI-produced visual content in a digital setting, the overall aim of RevealAI is to offer a user-friendly platform that would enable a user to upload their image in order to immediately acquire a trustworthy interpretation of the authenticity of that particular visual in the form of a confidence-based result. Unlike other traditional models, this platform would also offer a result that explains the reasoning behind it. The primary aims of this thesis were to examine current available AI-image detection systems in use, their limitations, and their improvement by developing a new tool that enhances explanation, usability, and accuracy. RevealAI has been designed and implemented in a modular fashion that differentiates between the front-end, back-end, and machine-learning components. The AI part of this system has been determined using a customized CNN model that has been developed in Python language using PyTorch library, performing a visual analysis of image artifacts. Technical metadata has been analyzed in a distinct analysis component, which would then help this system incorporate all types of evidence into a common prediction. The back-end part of this project has been developed using FastAPI, and PostgreSQL database, along with SQLAlchemy library has been used as a data management tool for storing data pertaining to user information, analysis data, vectors, explanations, and history. The frontend of RevealAI has been developed using React.js and Tailwind CSS programming languages. This provides a user-friendly interface that allows users to register, log in, upload their images for analysis, display results, and check their history of past analyses. User authentication has been protected using JWT tokens, which ensures that only authorized users are able to access their accounts. The raw images are temporarily stored for analysis and then immediately deleted, in line with data protection policies. RevealAI, with its capabilities of tech metadata analysis, visual CNN model analysis, and explanation reasoning, offers a comprehensive platform for differentiating between actual images and those produced by AI technology. RevealAI fills a gap that exists between advanced forensic analysis and lay users, as it creates an easily understandable platform for analyzing a societal challenge. The thesis has a total of 74 pages, 15 tables, and 26 figures. Keywords: artificial intelligence, image forensics, CNN model, metadata analysis, FastAPI, PyTorch, image evaluation.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Kostenko, Oleg,

Subjects

dc:subject × 1

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
eng

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Kostenko, Oleg,. A system for recognising the ai manipulation on photographs /. Institutional Repository of Vilnius University, 2026. https://repository.vu.lt/VU:ELABAETD277690607&prefLang=en_US