Skip navigation
Please use this identifier to cite or link to this item: https://repository.esi-sba.dz/jspui/handle/123456789/831
Title: Leveraging Transfer Learning and Domain Adaptation for Crop Mapping and Classification in Precision Agriculture
Authors: NAILI, NAda
Keywords: Crop Mapping
Transfer Learning
Domain Adaptation
Remote Sensing
Algeria
Issue Date: 2025
Abstract: Crop classiĄcation is crucial for precision agriculture and food security but remains challenging in data-scarce regions like North Africa. This thesis reviews recent advances in crop classiĄcation using remote sensing, focusing on machine learning, transfer learning, and domain adaptation techniques. It compares methods including Random Forests, CNNs, Transformers, and unsupervised adaptation strategies. Special attention is given to approaches that work under limited annotations and regional variability. Results show that while Transformer and adaptation methods are effective, their use in Algeria requires tailored strategies due to domain shifts and environmental complexity**** La classiĄcation des cultures est essentielle pour lŠagriculture de précision, mais elle reste difficile dans les régions pauvres en données, comme lŠAfrique du Nord. Ce mémoire passe en revue les méthodes récentes basées sur la télédétection, incluant lŠapprentissage automatique, lŠapprentissage par transfert et lŠadaptation de domaine. Les approches comparées (Random Forest, CNN, Transformers, UDA) montrent que les méthodes récentes sont prometteuses. Toutefois, leur application en Algérie nécessite des adaptations spéciĄques à cause des différences régionales et du manque dŠannotations.
Description: Supervisor :Mr. Meziane Iftene / Supervisor :Mr. Mohamed Lamine Larabi
URI: https://repository.esi-sba.dz/jspui/handle/123456789/831
Appears in Collections:Master

Files in This Item:
File Description SizeFormat 
Naili Nada Master Final Version .docx (1)-pages-1-1.pdf177,03 kBAdobe PDFView/Open
Show full item record


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.