PENGENALAN WAJAH SECARA REALTIME MENGGUNAKAN CONVOLUTIONAL NEURAL NETWORK PADA CITRA MULTI-FACE
Daftar Isi:
- This research is about realtime face recognition using convolutional neural network. In this research uses two convolutional neural network architectures, VGG16 and a simple convolutional neural network architecture consisting of two convolutional layers, one pooling layer, and two fully connected layers. The VGG16 architecture consists of 13 convolutional layers, 5 pooling layers, 2 fully connected layers. Offline testing is performed on AT & T Face Database and get an accuracy value of 95% on the Simple Convolutional Neural Network architecture and the accuracy obtained using VGG16 architecture is 98%. The test was also carried out offline and in realtime using data from 11 Informatics Engineering students at Sriwijaya University. Offline testing gets an accuracy of 99% using the Simple Convolutional Neural Network and an accuracy of 98% using the VGG16 architecture. For realtime testing accuracy value is 86% with an average respond time of 0.4 seconds using VGG16 architecture and 70% of accuracy with an average respond time of 0.02 seconds using the Simple Convolutional Neural Network architecture.