Pengenalan Wajah Penyandang Tunanetra Berbasis Deep Neural Network Untuk Papan Pengumuman Interaktif

Mukhairiq, Gusfatul (2024) Pengenalan Wajah Penyandang Tunanetra Berbasis Deep Neural Network Untuk Papan Pengumuman Interaktif. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022221058-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (36MB) | Request a copy

Abstract

Anak penyandang tunanetra memerlukan perhatian khusus, terlebih dalam rangka akses pendidikan mereka. Fasilitas sarana dan prasarana perlu disesuaikan sehingga anak dengan tuna netra dapat mendapatkan kualitas pendidikan yang sangat baik. Salah satu upaya dalam meningkatkan kualitas tersebut adalah dengan memberikan fasilitas interaktif dua arah untuk memudahkan pelajar tuna tetra mengakses informasi akademik di lingkungan sekolah. Penelitian ini memanfaatkan teknologi jaringan saraf tiruan untuk deteksi dan klasifikasi wajah sebagai tunanetra atau non- tunanetra dengan memanfaatkan algoritma MTCNN dan FaceNet. Pada proses pembuatan model penelitian ini menggunakan data untuk fase pelatihan 5141 , sementara 1285 digunakan untuk evaluasi model. Sistem yang dirancang telah menunjukkan kemampuan identifikasi wajah secara akurat, dengan tingkat akurasi mencapai 92.5% pada proses pelatihan model. Pengujian dilakukan pada 33 subjek berbeda dan menghasilkan keberhasilan yang cukup besar. Sistem juga menunjukkan performa yang kuat dalam pengujian komunikasi dua arah dengan rata-rata waktu respons antara 3.844 detik hingga 4.294 detik serta tingkat akurasi, keterbacaan, dan konsistensi respons mencapai 86.67%. Penerapan teknologi ini memberikan peningkatan inklusivitas dan pengalaman pendidikan anak-anak tunanetra dalam lingkungan akademik.
=================================================================================================================================
The visually impaired children need special attention, especially in order to access their education. Facilities and infrastructure need to be adjusted so that children with visual impairments can get an excellent quality of education. One of the efforts in improving this quality is to provide two-way interactive facilities to facilitate blind students to access academic information in the school environment. This research utilizes artificial neural network technology for face detection and classification as blind or non-visually impaired by utilizing the MTCNN and FaceNet algorithms. In the model building process, this study used 5141 data for the training phase, while 1285 were used for model evaluation. The designed system has demonstrated the ability to accurately identify faces, with the accuracy rate reaching 92.5% in the model training process. Testing was conducted on 33 different subjects and resulted in considerable success. The system also showed strong performance in two-way communication tests with an average response time between 3.844 seconds to 4.294 seconds and response accuracy, readability and consistency rates reaching 86.67%. The application of this technology provides improved inclusivity and educational experience for visually impaired children in an academic setting.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	deep learning, facenet, tunanetra, deep learning, MTCNN, visually impaired
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Gusfatul Mukhairiq
Date Deposited:	26 Jul 2024 04:52
Last Modified:	26 Jul 2024 04:52
URI:	http://repository.its.ac.id/id/eprint/109432

Actions (login required)

View Item