Comparison of Convolutional Neural Networks Model Using Different Optimizers for Image Classification
I Ketut Adi Wirayasa, Handri Santoso, Eko Indrajit · International Journal of Sciences: Basic and Applied Research · 2021
Face detection technology and image classification are widely used in several industries that help humans in obtaining information and other related matters. In this paper, the utilization of the Computer Vision system uses the Convolutional Neural Network (CNN) algorithm to classify images by distinguishing the gender of the detected object. Architectural model through transfer learning by experimenting with three pre-trained models, namely VGG-16, Inception-V3, and MobileNet-V2 to determine the best architecture by using Optimizer Adam and RMSProp. To produce the best model and performance, experiments were carried out using several modules such as the data augmentation module and the re-indexing module. The Inception-V3 model got the best results in predicting Gender from the image with an accuracy and loss validation value of 0.9350, 0.1550, compared to VGG-16 and MobleNet-V2 with values 0.9320, 0.1660, and 0.8760, 0.3000.