Generating Arabic Calligraphy using Generative Adversarial Networks (GANs) / Hadj Azze, Yousra Chahinez  

Public ISBD Titre : Generating Arabic Calligraphy using Generative Adversarial Networks (GANs) Type de document : texte imprimé Auteurs : Hadj Azze, Yousra Chahinez, Auteur ; Abdelouahab Moussaoui, Directeur de thèse Editeur : Setif:UFA Année de publication : 2021 Importance : 1 vol (75 f .) Format : 29 cm Langues : Français (fre) Catégories : Thèses & Mémoires:Informatique Mots-clés : Arabic calligraphy Convolutional Neural Nets Index. décimale : 004 Informatique Résumé : Arabic Calligraphy is a unique art and an important manifestation of Arabic culture, especially because it is the first script that was used to write the Quran. It is used in many applications such as glorify and beautify God’s names, conceive and plan out interior and exterior homes decors and mosques, coin design, stories design in social media, etc. Despite the importance of Arabic Calligraphy and its potential usefulness for many applications, there are no studies on the automatic generation of calligraphy works, due to the complexity of the Arabic calligraphy and the lack of training data. Generative Adversarial Network (GAN) has made a breakthrough and great success in many research areas in computer vision. Different GANs generate different outputs. In this work, we apply GANs to generate Arabic calligraphy words. Conditional GAN for the image to image translation (Pix2Pix) and Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (cycleGAN) are used. We treat the synthesis process as an image-to-image translation task in order to synthesize Arabic Calligraphy images with specified style (Reqaa, Farisi, and Diwani) from source font (ex. Arial) images. Then, the results of the generated images are evaluated using native-Arabic human and Frechet Inception Distance (FID). The experiments shows that the images generated using CycleGAN are better than Pix2Pix, which proved by the FID score. The results often contains some erroneous and missing strokes. Côte titre : MAI/0473 En ligne : https://drive.google.com/file/d/1zMEQa5SpPZIhmJiEUZXfQwI91k8Vz1Lo/view?usp=shari [...] Format de la ressource électronique : pdf Generating Arabic Calligraphy using Generative Adversarial Networks (GANs) [texte imprimé] / Hadj Azze, Yousra Chahinez, Auteur ; Abdelouahab Moussaoui, Directeur de thèse . - [S.l.] : Setif:UFA, 2021 . - 1 vol (75 f .) ; 29 cm. Langues : Français (fre) Catégories : Thèses & Mémoires:Informatique Mots-clés : Arabic calligraphy Convolutional Neural Nets Index. décimale : 004 Informatique Résumé : Arabic Calligraphy is a unique art and an important manifestation of Arabic culture, especially because it is the first script that was used to write the Quran. It is used in many applications such as glorify and beautify God’s names, conceive and plan out interior and exterior homes decors and mosques, coin design, stories design in social media, etc. Despite the importance of Arabic Calligraphy and its potential usefulness for many applications, there are no studies on the automatic generation of calligraphy works, due to the complexity of the Arabic calligraphy and the lack of training data. Generative Adversarial Network (GAN) has made a breakthrough and great success in many research areas in computer vision. Different GANs generate different outputs. In this work, we apply GANs to generate Arabic calligraphy words. Conditional GAN for the image to image translation (Pix2Pix) and Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks (cycleGAN) are used. We treat the synthesis process as an image-to-image translation task in order to synthesize Arabic Calligraphy images with specified style (Reqaa, Farisi, and Diwani) from source font (ex. Arial) images. Then, the results of the generated images are evaluated using native-Arabic human and Frechet Inception Distance (FID). The experiments shows that the images generated using CycleGAN are better than Pix2Pix, which proved by the FID score. The results often contains some erroneous and missing strokes. Côte titre : MAI/0473 En ligne : https://drive.google.com/file/d/1zMEQa5SpPZIhmJiEUZXfQwI91k8Vz1Lo/view?usp=shari [...] Format de la ressource électronique : pdf

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
MAI/0473	MAI/0473	Mémoire	Bibliothéque des sciences	Anglais	Disponible Disponible

---

1 (1 - 1 / 1)