University Sétif 1 FERHAT ABBAS Faculty of Sciences
Détail de l'auteur
Auteur sarra Laidani |
Documents disponibles écrits par cet auteur
Ajouter le résultat dans votre panier Affiner la recherche
Titre : Image Enhancement with Deep Learning : Restoration of historical documents using gans Type de document : texte imprimé Auteurs : sarra Laidani, Auteur ; Abderrahim Zouaid, Auteur ; Safia Zazoua, Directeur de thèse Année de publication : 2022 Importance : 1 vol (91 f .) Format : 29cm Langues : Français (fre) Catégories : Thèses & Mémoires:Informatique Mots-clés : Document enhancement
Pix2PixIndex. décimale : 004 Informatique Résumé :
Documents often demonstrate different degradation conditions, making them difficult to read,
ancient historical papers. Therefore, in this project, we suggest a practical image-to-image
translation framework to highlight the endeavor to the restorative historical documents that use
the conditional GANs (cGANs) to restore many historical degraded documents. Moreover, despite
the lack of resources, we sought to produce an improved version of the degraded document with
high quality.
In addition to using Pix2Pix and CycleGAN model, we compare how our approach provides
consistent improvements based on these two models, demonstrating the capacity to restore a
degraded printed document image and handwritten document to a more satisfactory and readable
state.
we used datasets from the International Document Image Binarization Contest (DIBCO)
from 2009 until DIBCO 2020 and also Handwritten document image binarization (H-DIBCO)
from 2010 until 2020.Côte titre : MAI/0620 En ligne : https://drive.google.com/file/d/1nsQako6vKI1FHSsOkFAJCtaWxzKarVie/view?usp=share [...] Format de la ressource électronique : Image Enhancement with Deep Learning : Restoration of historical documents using gans [texte imprimé] / sarra Laidani, Auteur ; Abderrahim Zouaid, Auteur ; Safia Zazoua, Directeur de thèse . - 2022 . - 1 vol (91 f .) ; 29cm.
Langues : Français (fre)
Catégories : Thèses & Mémoires:Informatique Mots-clés : Document enhancement
Pix2PixIndex. décimale : 004 Informatique Résumé :
Documents often demonstrate different degradation conditions, making them difficult to read,
ancient historical papers. Therefore, in this project, we suggest a practical image-to-image
translation framework to highlight the endeavor to the restorative historical documents that use
the conditional GANs (cGANs) to restore many historical degraded documents. Moreover, despite
the lack of resources, we sought to produce an improved version of the degraded document with
high quality.
In addition to using Pix2Pix and CycleGAN model, we compare how our approach provides
consistent improvements based on these two models, demonstrating the capacity to restore a
degraded printed document image and handwritten document to a more satisfactory and readable
state.
we used datasets from the International Document Image Binarization Contest (DIBCO)
from 2009 until DIBCO 2020 and also Handwritten document image binarization (H-DIBCO)
from 2010 until 2020.Côte titre : MAI/0620 En ligne : https://drive.google.com/file/d/1nsQako6vKI1FHSsOkFAJCtaWxzKarVie/view?usp=share [...] Format de la ressource électronique : Exemplaires (1)
Code-barres Cote Support Localisation Section Disponibilité MAI/0620 MAI/0620 Mémoire Bibliothéque des sciences Anglais Disponible
Disponible