Analisis Metode Konvolusi dengan Pendekatan Jaringan Saraf Tiruan untuk Mendeteksi Keaslian Citra Digital

Abstract

Digital images are becoming increasingly important in toda y's digital era, but with technological developments, the problem of authenticity of digital images also arises. Therefore, this swdy aims to analyze the convolution method with an artificial neural network approach to detect the authenticity of digital images. This study uses a11 artificial neural network approach using a convolution architecture which has been proven effective in image processing. The training data consists of original digital images a11d modified digital images obtained from kaggle.com. The training process is carried out byoptimizing the network weights andbiases through the backpropagation algorithm. The results showed that the convolution method with the artificial neural network approach obtained an accuracy of 72.21% in detecting the authemicity of digital images with the VGG16 parameter as a pretrained model for extracting features with 128 neurons in the hidden layer, the ReLUactivation fu11ction in the hidden layer a11d a /eami 11g rate of 0.05.