RONDIYANTO, RONDIYANTO (2022) IMPLEMENTASI FRAMEWORK TENSORFLOW OBJECT DETECTION DALAM MENDETEKSI GESTUR BAHASA ISYARAT. Undergraduate thesis, Universitas Stikubank.
![[thumbnail of HALAMAN JUDUL]](/style/images/fileicons/application_pdf.png) |
PDF (HALAMAN JUDUL)
Download (641kB) |
|
PDF (ABSTRAK)
Download (83kB) |
|
PDF (BAB I)
Download (93kB) |
|
PDF (BAB II)
Restricted to Repository staff only Download (157kB) |
|
PDF (BAB III)
Restricted to Repository staff only Download (1MB) |
|
PDF (BAB IV)
Restricted to Repository staff only Download (921kB) |
|
PDF (BAB V)
Restricted to Repository staff only Download (105kB) |
|
PDF (DAFTAR PUSTAKA)
Download (78kB) |
|
PDF (LAMPIRAN)
Restricted to Repository staff only Download (1MB) |
Abstract
Penyandang tunarungu adalah orang yang pendengarannya tidak berfungsi dengan baik sehingga membutuhkan alat bantu khusus untuk berkomunikasi. Bahasa Isyarat Indonesia (BISINDO) merupakan salah satu alat bantu yang digunakan tunarungu dalam menyampaikan dan memahami pesan. Namun dengan jumlah kosakata BISINDO yang sangat banyak perlu adanya teknologi untuk membantu proses klasifikasi dan mendeteksi gestur kosakata BISINDO agar mudah dimengerti terutama bagi orang awam. Penelitian ini menggunakan Framework Tensorflow Object Detection API dengan metode Convolutional neural network (CNN) dan arsitektur SSD MobileNet V2 FPNLite 320x320 untuk melakukan klasifikasi dan deteksi gestur kosakata BISINDO. Kosakata terdiri dari 7 gestur yang diambil dari nama-nama “Hari”, dengan total data sebanyak 1654 citra. Setelah proses training model yang dilakukan, menghasilkan prediksi akurasi rata-rata sebesar 97,14% pada epoch ke 4.000. Deaf people are people whose hearing is not functioning properly so they need special aids to communicate. Indonesian Sign Language (BISINDO) is one of the tools used by the deaf in conveying and understanding messages. However, with a large number of BISINDO vocabularies, technology is needed to assist the classification process and detect BISINDO vocabulary gestures so that they are easy to understand, especially for ordinary people. This study uses the Tensorflow Object Detection API Framework with the Convolutional neural network (CNN) method and the MobileNet V2 FPNLite 320x320 SSD architecture to classify and detect BISINDO vocabulary gestures. Vocabulary consists of 7 gestures taken from the name "Day", with a total data of 1654 images. After the model training process is carried out, it produces an average accuracy prediction of 97.14% at the 4,000th epoch.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Additional Information: | SKR.I.05.01.2198 NIM.17.01.53.0056 |
| Uncontrolled Keywords: | BISINDO, Deteksi Objek, Convolutional Neural Network, Tensorflow object detection BISINDO, Deteksi Objek, Convolutional Neural Network, Tensorflow object detection |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science |
| Faculty / Institution: | Fakultas Teknologi Informasi > Program Studi Teknik Informatika |
| Depositing User: | Teteh Hayati |
| Date Deposited: | 27 Sep 2022 02:53 |
| Last Modified: | 27 Sep 2022 02:53 |
| URI: | https://eprints.unisbank.ac.id/id/eprint/8718 |
Actions (login required)
 |
View Item |
