Augmentasi Data Citra Ekspresi Wajah Berdasarkan Transformasi Geometris Dan Filtering Pada Sistem Pengenalan Ekspresi Wajah.

Sofyantoro, Rahma (2022) Augmentasi Data Citra Ekspresi Wajah Berdasarkan Transformasi Geometris Dan Filtering Pada Sistem Pengenalan Ekspresi Wajah. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111640000117-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (22MB)

Abstract

Ekpresi wajah menjadi bagian penting dalam mengetahui bagaimana menilai perasaan manusia. Teknologi pengenalan eskpresi wajah berguna untuk mengetahui secara otomatis bagaimana reaksi orang-orang terhadap fenomena tertentu yang dihadapnya. Salah satu metode yang dapat digunakan untuk pengenalan ekspresi wajah adalah Deep learning. Salah satu kendala dalam metode ini adalah keterbatasan data training yang nantinya akan mengurangi peiforma dari model yang dihasilkan .Tugas akhir ini mengimplementasikan sistem pengenalan ekspresi wajah menggunakan Convolutional Neural Network (CNN) Pretrained VGG-16 dengan data traning berupa data citra ekpresi wajah yang telah dilakukan augmentasi. Augmentasi dilakukan dengan melakukan random transformasi geometris terhadap dataset citra ekpresi wajah dan memberikan filtering terhadap data tersebut. Hasil dari Augmentasi terdiri dari 8 variasi dataset transormasi geometris dan 3 variasi dataset filtering. Dataset hasil augmentasi ini yang digunakan sebagai data training model. Hasil dari pengujian pada data testing CK + di dapatkan bahwa model dengan data training asli (tanpa augmentasi) ditambah hasil augmentasi transformasi geometris translasi mendapatkan akurasi testing sebesar 0.962 lebih besar dari pada data asli yang hanya mendapatkan 0.9348. Pada data testing IMED, model dengan data training asli ditambah hasil augmentasi transformasi geometris translasi dan rotasi mendapatkan akurasi testing sebesar 0.5308 lebih besar dari pada data asli yang hanya mendapatkan 0.5231.
==================================================================================================================================
Facial expressions are an important part of knowing how to judge human feelings. Facial expression recognition technology is useful for automatically knowing how people react to certain phenomena they face. One method that can be used for facial expression recognition is deep learning. One of the obstacles in this method is limited training data which will reduce the performance of resulting model. This final project implements a facial expression recognition system using the Convolutional Neural Network (CNN) Pre-trained VGG-16 with training data in the form of facial expression image data that augmentation has been performed. Augmentation is done by performing random geometric transformations on the facial expression image dataset and providing filtering to the data. The results of the Augmentation consist of 8 variations of the geometric transformation dataset and 3 variations of the filtering dataset. The augmented dataset is used as the training model data. The results of the test on the CK + testing data show that the model with the original training data (without augmentation) plus the translational geometric transformation augmentation results in a testing accuracy of 0.962, which is greater than the original data which only gets 0.9348. In the IMED testing data, the model with the original training data plus the results of the augmentation of the translational and rotational geometric transformations gets a testing accuracy of0.5308 which is greater than the original data which only gets 0.5231.

Item Type:	Thesis (Other)
Additional Information:	RSIf 006.42 Sof a-1 2022
Uncontrolled Keywords:	Ekpresi wajah, Deep Learning, Convolutional Neural Network,Augmentasi. Facial expression, Deep Learning, Convolutional Neural Network, Augmentation.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	25 May 2026 01:46
Last Modified:	25 May 2026 01:46
URI:	http://repository.its.ac.id/id/eprint/133369

Actions (login required)

View Item