Please use this identifier to cite or link to this item:
http://dspace.ensta.edu.dz/jspui/handle/123456789/649Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.author | BENYOUCEF, Meriem | - |
| dc.contributor.author | CHOUAF, Seloua (Directeur de thèse) | - |
| dc.date.accessioned | 2026-11-12T13:21:47Z | - |
| dc.date.available | 2026-11-12T13:21:47Z | - |
| dc.date.issued | 2022 | - |
| dc.identifier.uri | http://dspace.ensta.edu.dz/jspui/handle/123456789/649 | - |
| dc.description | Projet de fin d’étude d'ingéniorat: Systèmes de Télécommunications Et Réseaux: Alger: Ecole Nationale Supérieure des Technologie Avancées (ex ENST): 2022. | en_US |
| dc.description.abstract | This dissertation aims to contribute to create an automated description model based on deep learning architectures to generate human-like textual descriptions of images. This work combines different convolutional neural networks (CNN) for the feature extraction part, and long-term memory (LSTM) for the textual generation part. Testing these models through a series of experiments yielded different results in terms of loss, accuracy and natural language evaluation metrics. | en_US |
| dc.language.iso | en | en_US |
| dc.publisher | ENST | en_US |
| dc.subject | Deep learning | en_US |
| dc.subject | Convolutional neural networks (CNN) | en_US |
| dc.subject | Recurrent neural networks (RNN) | en_US |
| dc.subject | Long Short Time Memory (LSTM) | en_US |
| dc.subject | Natural Language Processing (NLP) | en_US |
| dc.subject | Automated Image Description | en_US |
| dc.title | Automated Image Description Based on Multimodal Deep Learning | en_US |
| dc.type | Thesis | en_US |
| Appears in Collections: | PFE- Systèmes de Télécommunications et Réseaux | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| pfe.2022.STR.BENYOUCEF.Meriem.pdf | Projet d'ingéniorat | 4.53 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.