Laksono, Fathullah Auzan Setyo (2022) Estimasi Umur, Gender Dan Ras Menggunakan Convolutional Neural Network Berbasis Citra Wajah. Other thesis, Institut Teknologi Sepuluh Nopember.
|
Text
07211840000053-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (20MB) | Request a copy |
Abstract
Identifikasi umur, gender dan ras dapat sangat berguna dalam banyak pengimplementasian ilmu seperti pada pengamatan visual, diagnosa medis, sistem interaksi komputer manusia, biometric, pengumpulan informasi, penegakan hukum, pemasaran dan banyak lainnya. Dimana sebagian besar data mengenai fitur wajah tersebut masih diambil secara manual melalui survei ataupun pengamatan pada banyak individu. Oleh karena itu perlu dibuat suatu sistem yang dapat mengestimasi umur, gender dan ras untuk mempermudah pengumpulan data. Sistem yang dibuat akan menggunakan Convolutional Neural Network untuk melakukan estimasi umur, gender dan ras dari citra wajah. Dengan menggunakan transfer learning pada arsitektur ResNet50, ResNet50V2, VGG16, VGG19, Inception V3, InceptionResNetV2, XCeption, EfficienNetB0 dan pembuatan arsitektur CNN sederhana. Setelah dilakukan training, validation dan testing, didapatkan akurasi paling tinggi menggunakan model dengan arsitektur ResNet50V2. Dengan nilai akurasi klasifikasi umur 0,6, akurasi klasifikasi gender 0,88 dan akurasi klasifikasi ras 0,72. Model yang sudah disimpan diimplementasikan pada prototype webapp yang akan mendeteksi citra wajah dari gambar dan mengestimasikan umur, gender dan ras dengan waktu sebanyak 2,65 detik. Diharapkan dalam penelitian ini, kedepannya dapat dikembangkan suatu sistem yang lebih kompleks untuk mengumpulkan data umur, gender dan ras dengan lebih mudah dan efisien.
==============================================================================================================================
Identification of age, gender and race can be very useful in many science implementations such as in visual observation, medical diagnosis, human computer interaction systems, biometrics, information gathering, law enforcement, marketing and many others. Where most of the data regarding these facial features are still taken manually through surveys or observations on many individuals. Therefore, it is necessary to create a system that can estimate age, gender and race to facilitate data collection. The system created will use the Convolutional Neural Network to estimate age, gender and race from facial images. By using transfer learning on the ResNet50, ResNet50V2, VGG16, VGG19, Inception V3, InceptionResNetV2, XCeption, EfficienNetB0 architectures and building a simple CNN architecture. After training, validation and testing, the highest accuracy was obtained using a model with the ResNet50V2 architecture. With an age classification accuracy value of 0.6, gender classification accuracy 0.88 and race classification accuracy 0.72. The saved model is implemented on a webapp prototype that will detect facial images from images and estimate age, gender and race in 2.65 seconds. It is hoped that in this research, in the future a more complex system can be developed to collect data on age, gender and race more easily and efficiently.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSKom 006.4 Lak e-1 2022 |
| Uncontrolled Keywords: | CNN, Umur, Gender, Ras, Citra, Wajah. CNN, Age, Gender, Race, Facial. |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis |
| Depositing User: | Mr. Marsudiyana - |
| Date Deposited: | 17 Jun 2026 03:38 |
| Last Modified: | 17 Jun 2026 03:38 |
| URI: | http://repository.its.ac.id/id/eprint/133848 |
Actions (login required)
![]() |
View Item |
