Model Validasi Gambar Target Dengan Modifikasi Algoritma Deep Neural Network pada Sensor Gambar

Abstract

Image validation is always necessary to ensure that detected images are accurate. The higher the object image validation value, the better the object detection performance. The development of Artificial Neural Networks (ANN) has reached the stage where the concept of Deep Neural Networks (DNN) has emerged as a promising evolution of ANN. DNN is one of the algorithms in Deep Learning that can efficiently solve complex problems with large datasets. Therefore, it is essential to study the best model to be applied to DNN to achieve optimal target image validation performance. This research discusses the validation of target images by modifying the DNN algorithm. The hidden layers examined include 3, 4, 5, and 6 hidden layers. In addition to the number of hidden layers, the study also considers variations in activation functions, batch sizes, and the number of neurons. The results show that the best performance is achieved with a model using 3 hidden layers, a new activation function and Sigmoid, a batch size of 64, and the number of neurons set at 256, 128, and 64, respectively. This model is capable of achieving realtime target image validation accuracy of up to 95.46%.