Shadri, Muhamad (2025) Deteksi Gambar Yang Dihasilkan Ai Menggunakan Algoritma Convolutional Neural Network(cnn). Other thesis, Universitas Islam Riau.

Text 203510002.pdf - Submitted Version Restricted to Registered users only Download (6MB) | Request a copy

Abstract

The detection of AI-generated images has become a crucial challenge in today's digital era, especially in ensuring the authenticity of visual content. This study aims to develop an image classification system to distinguish AI-generated and non-AI-generated images using Convolutional Neural Networks (CNN) with various model architectures, including MobileNet, MobileNetV2, ResNet50, and a Custom CNN. The dataset consists of 16,000 images, equally divided into 8,000 AI-generated and 8,000 non-AI-generated images. Before training, the dataset was split into 80% for training, 10% for validation, and 10% for testing. The experiments were conducted using the Adam and RMSprop optimizers, each with learning rate variations of 0.001 and 0.0001, and trained for 50 and 100 epochs. The results indicate that the MobileNetV2 model, optimized with Adam, a learning rate of 0.001, and 100 epochs, achieved the best performance with a training accuracy of 92.87% and a testing accuracy of 88%. Model evaluation shows that MobileNetV2 effectively detects AI-generated images with a precision of 0.81, recall of 0.96, and an F1-score of 0.88. Meanwhile, for the non-AIgenerated class, the model achieved a precision of 0.95, recall of 0.77, and an F1- score of 0.85. With its high performance and better efficiency compared to other models, MobileNetV2 proves to be an optimal choice for AI and non-AI image classification tasks.

Actions (login required)

View Item