Putra, Naufal Aurelio (2022) Segmentasi Area Wajah Menggunakan Cnn. Other thesis, Institut Teknologi Sepuluh Nopember.
|
Text
07211840000017-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (18MB) |
Abstract
Segmentasi area wajah merupakan bidang visi komputer yang melakukan segmentasi piksel pada gambar wajah menjadi dengan label tertentu sesuai dengan semantiknya. Segmentasi area wajah mengambil peran penting dalam berbagai implementasi analisis wajah manusia. Meskipun demikian, sebagian besar metode segmentasi area wajah yang telah ada masih kurang mampu dalam menangani skenario input gambar yang kompleks. Masalah ini dapat diselesaikan dengan menggunakan metode deep learning khususnya Convolutional Neural Network (CNN) yang dapat mengabstraksi fitur pada gambar yang berhubungan dengan data spasial dengan baik dan dapat beradaptasi dengan input gambar yang kompleks. Saat ini, penggunaan metode CNN untuk segmentasi area wajah masih menghadapi tantangan dalam implementasinya dikarenakan masalah jumlah dan variasi data serta model CNN yang bervariasi sehingga diperlukan eksplorasi lebih jauh untuk mengetahui performa model tersebut. Dengan demikian, penelitian tugas akhir ini berfokus pada penggunaan CNN untuk segmentasi area wajah dengan tujuan untuk dapat membagi piksel gambar wajah menjadi segmen tertentu secara akurat. Pada penelitian ini, digunakan arsitektur Bilateral Segmentation Network (BiSeNet) dan dataset gambar wajah dengan komposisi data dan kondisi gambar wajah yang bervariasi. Berdasarkan hasil testing akhir, diperoleh performa model yang akurat dengan penilaian Mean IOU 80.34% pada dataset benchmark.
==============================================================================================================================
Face area segmentation is a field of computer vision that performs pixel-wise segmentation on face images into certain labels according to their semantics. Face area segmentation has an important role in various implementation of human face analysis. However, most of the existing face area segmentation methods are difficult to handle complex image input scenarios. This problem can be solved by implementing deep learning method, especially Convolutional Neural Network (CNN) which enables to abstract features related to spatial data in images properly and adapt to complex image inputs. Currently, the utilization of CNN method in face area segmentation still faces challenges due to the problem of the data variability and CNN model variability so that further exploration is needed in order to determine the performance of the model. Hence, this final project focuses on the utilization of CNN for face area segmentation with the aim of being able to accurately segment pixels in the face area. In this final project, the Bilateral Segmentation Network (BiSeNet) architecture will be used with facial image dataset that has a variety of composition and conditions. Based on testing result, an accurate model performance is obtained with a Mean IOU rating of 80.34% on the benchmark dataset.
| Item Type: | Thesis (Other) |
|---|---|
| Additional Information: | RSKom 006.42 Put s-1 2022 |
| Uncontrolled Keywords: | Face Area Segmentation, Convolutional Neural Network, Deep Learning. Segmentasi Area Wajah, Convolutional Neural Network, Deep Learning. |
| 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 01:01 |
| Last Modified: | 17 Jun 2026 01:01 |
| URI: | http://repository.its.ac.id/id/eprint/133827 |
Actions (login required)
![]() |
View Item |
