Micro-Expression Recognition Using VGG19 Convolutional Neural Network Architecture And Random Forest

Micro-expression is a hidden expression of human. This expression occurs in a very short time and very difficult to see or detect in a daily conversation. A system that developed in this research is a system that capable of analyzing micro-expression and show the result performance. The method used is by using Convolutional Neural Network (CNN) with Visual Geometry Group - 19 (VGG19) architecture as feature extractor and random forest as classifier. Architecture CNN VGG19 selected because many case in image classification, CNN always have good performance. The research use SMIC-Cropped database with the distribution of percentage between training:validation of 50:50, 55:45, 60:40, 65:35, 70:30, 75:25, 80:20, 85:15, and 90:10. This research obtained accuracy in the range of 98% - 100%. The distribution percentage of the database which have best performance is obtained when the split is 90:10 with result 99% - 100% accuracy, 99%-100% precision, 100% recall, and 100% f1-score.