Amalia, Aifa Nur (2019) Sintesis teks ke gambar untuk pembuatan gambar digital menggunakan algoritma Generative Adversarial Networks. Diploma thesis, UIN Sunan Gunung Djati Bandung.
|
Text (COVER)
1_cover.pdf Download (90kB) | Preview |
|
|
Text (ABSTRAK)
2_abstrak.pdf Download (43kB) | Preview |
|
|
Text (DAFTAR ISI)
3_daftarisi.pdf Download (137kB) | Preview |
|
|
Text (BAB I)
4_bab1.pdf Download (90kB) | Preview |
|
Text (BAB II)
5_bab2.pdf Restricted to Registered users only Download (587kB) | Request a copy |
||
Text (BAB III)
6_bab3.pdf Restricted to Registered users only Download (826kB) | Request a copy |
||
Text (BAB IV)
7_bab4.pdf Restricted to Registered users only Download (2MB) | Request a copy |
||
Text (BAB V)
8_bab5.pdf Restricted to Registered users only Download (108kB) | Request a copy |
||
Text (DAFTAR PUSTAKA)
9_daftarpustaka.pdf Restricted to Registered users only Download (176kB) | Request a copy |
Abstract
INDONESIA Deep learning merupakan serangkaian metode yang terdiri dari multi-layer neural networks sederhana yang memungkinkan sistem untuk menyelesaikan masalah-masalah yang membutuhkan pemahaman yang setara dengan kecerdasan manusia. Salah satu penerapan deep learning yang banyak dikaji adalah sintesis teks ke gambar. Sintesis teks ke gambar pada dasarnya menerjemahkan teks atau kalimat yang diinginkan ke dalam sebuah gambar yang digenerasi oleh komputer secara otomatis. Pada penelitian sebelumnya, algoritma Generative Adversarial Networks (GAN) digunakan untuk melakukan sintesis berupa gambar foto realistik, namun belum banyak digunakan untuk generasi gambar yang mengandung nilai seni seperti motif batik. Pada penelitian ini, dibangun sebuah sistem yang menerapkan algoritma GAN untuk melakukan sintesis teks ke gambar motif batik menggunakan bahasa pemrograman Python. Proses sintesis ini terdiri dari tahap data loading, tahap training inference dengan menggunakan Convolutional Neural Networks (CNN) serta Recurrent Neural Networks (RNN), dan tahap training yang terbentuk dari fungsi generator dan discriminator. Hasil penelitian menunjukkan bahwa gambar sintesis memiliki tingkat kemiripan tertinggi, yaitu 51.43% untuk Step I serta 53.81% untuk Step II dengan learning rate sebesar 0.00000625. Sedangkan, tingkat akurasi precision and recall masing-masing sebesar 0.71 dan 0.78 untuk Step I serta 0.83 dan 0.86 untuk Step II. ENGLISH Deep learning is a series of methods consisting of simple multi-layer neural networks that allows the system to solve problems that are equivalent to human intelligence which is one of the applications of deep learning that is widely studied in text-to-image synthesis. Text-to-image synthesis basically translates the desired text or sentence into an image generated by a computer automatically. In previous studies, Generative Adversarial Networks (GAN) algorithms were used to synthesize realistic photo images, but not many of these were used for the generation of images containing artistic values such as batik pattern. In this study, a system was developed that applies the GAN algorithm to synthesize text to batik pattern images using the Python programming language. This synthesis process consists of loading data stages, training inference stages using Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), and training stages formed from the function of the generator and discriminator. The results of this study indicate that synthetic images have the highest level of similarity, namely 51.43% for Step I and 53.81% for Step II with the learning rate of 0.00000625. Meanwhile, the accuracy of precision and recall is 0.71 and 0.78 for Step I and 0.83 and 0.86 for Step II.
Item Type: | Thesis (Diploma) |
---|---|
Uncontrolled Keywords: | Deep Learning; Generative Adversarial Networks; Convolutional Neural Networks; Recurrent Neural Networks; Precision and Recall; Python; |
Subjects: | Data Processing, Computer Science > Computer Science Education Data Processing, Computer Science > Computer and Human Special Computer Methods > Artificial Intelligence |
Divisions: | Fakultas Sains dan Teknologi > Program Studi Teknik Informatika |
Depositing User: | Aifa Nur Amalia |
Date Deposited: | 28 May 2019 07:18 |
Last Modified: | 28 May 2019 07:18 |
URI: | https://digilib.uinsgd.ac.id/id/eprint/20683 |
Actions (login required)
View Item |