DSpace Collection:
https://dspace.univ-ouargla.dz/jspui/handle/123456789/243
2024-03-29T09:58:26ZOligonychus detection Using YOLO ALGORITHM
https://dspace.univ-ouargla.dz/jspui/handle/123456789/35051
Titre: Oligonychus detection Using YOLO ALGORITHM
Auteur(s): BELHADJ, Mourad; Benali, Mohammed Anis; Benbelgacem, Yacine Abdelatif
Résumé: Olygonichus is a widespread crop disease causing concern for farmers globally. Traditional methods have proven ineffective, but recent advances in AI and drone technology offer promise in controlling it. This discussion focuses on employing AI-powered drones for Olygonichus disease management, potentially boosting crop yields and food security. The YOLO Algorithm is used for accurate disease detection. With this algorithm, drones can both identify and treat Olygonichus disease: 1. Identification: Trained on images of affected crops, the YOLO algorithm enables drones to swiftly recognize Olygonichus disease. It marks affected areas by drawing boxes around them. 2. Treatment: Equipped with specialized tools, drones can administer targeted treatments to afflicted crops. For instance, they can carry sprayers to apply pesticides directly to affected plants. This precise targeting reduces the need for broad treatment and minimizes harm to unaffected crops. By merging the YOLO algorithm’s identification capabilities with the treatment delivery potential of drones, farmers can efficiently manage Olygonichus disease. This targeted approach safeguards overall crop yield and enhances food security2023-01-01T00:00:00ZCOVID -19 SCREENING BASED ON SUPERVIED DEEP LEARNING
https://dspace.univ-ouargla.dz/jspui/handle/123456789/35045
Titre: COVID -19 SCREENING BASED ON SUPERVIED DEEP LEARNING
Auteur(s): AIADI, Oussama; FASSOULI, FAYCAL
Résumé: CT-scans images are helpful for detecting COVID-19. In this thesis, we are interested by investigating the performance of deep networks with CT-scan images and with detecting COVID-19. We used Convolution deep networks and we will provide insights on the performance of this deep networks. also, compare the performance of this networks and using ensemble learning we try to combine different models. The proposed approach is three-fold and comprises three stages which are training, training with data augmentation then ensemble learning. The first stage is done to train our models with CT-scan images and see their performance. While the second stage is dedicated to see how models will perform using data augmentation technique on CT-scan images. The last stage is using ensemble learning and combine the results of models to increase the accuracy. We conduct experiment on COVID-19CT dataset. With accuracy as performance metric. Experimental results reveal that model without data augmentation are more performing than models with it. VGG16 model is better performing than ResNet50.2023-01-01T00:00:00ZFIRST ORDER OPTIMIZATION METHODS FOR DEEP LEARNING.
https://dspace.univ-ouargla.dz/jspui/handle/123456789/35043
Titre: FIRST ORDER OPTIMIZATION METHODS FOR DEEP LEARNING.
Auteur(s): BOUANANE, KHADRA; DOKKAR, BASMA; MEDDOUR, BOUTHAYNA
Résumé: Deep learning has emerged as a transformative technology in various domains, ranging from computer vision to natural language processing. The success of deep learning models heavily relies on effective optimization algorithms. In this thesis, two main contributions are presented. In Contribution 1, which is a two-fold comparative study, we first explore the impact of various first-order optimization techniques on the learning process of U-Net for the task of Change Detection. Namely, Gradient descent with Momentum (Momentum GD), Nesterov Accelerated Gradient (NAG), Adaptive Gradient (AdaGrad), Root Mean Square Propagation optimizer (RMSProp), and the adaptive moment estimation optimizer (Adam). The results show that RMSProp, NAG, and AdaGrad reached the highest validationaccuracies: 0.976,0.978,and0.979with10−2, 10−3,and10−4 respectively,whileAdam was the fastest to converge and scored the lowest validation loss. Moreover, Adam scored the highest precision and F1 score across all learning rate values with 0.491 and 0.376 respectively. Nevertheless, we noticed that Adam’s performance could be significantly influenced by the data sparsity. In light of this hypothesis, the second part of Contribution 1 investigates the impact of sparsity on the performance of Adam optimizer. We compare different sparsity-level models, U-Net, DenseU-Net, and DenseNet using Adam optimizer for BCE and focal Tversky losses, on dense and sparse datasets for three ML tasks: Change detection, image segmentation, and object recognition. According to the obtained results, the Adam optimizer seems to be more sensitive to the model than the data sparsity. In Contribution 2, we propose a new method that aims to improve Adam’s performance. In this approach, we combine a simulated annealing strategy with a dynamic learning rate
iii
IV
to overcome the generalization gap which characterizes adaptive methods. We assess the several variants of the proposed approach compared to Adam, stochastic Gradient Descent, and Adabound. For this purpose, a simple 3-layer CNN is trained on two datasets MNIST and CIFAR-10.2023-01-01T00:00:00ZSécurité des échanges par stéganographie d’image numérique
https://dspace.univ-ouargla.dz/jspui/handle/123456789/35041
Titre: Sécurité des échanges par stéganographie d’image numérique
Auteur(s): Khaldi, Amine; Benhelal, Radia; Benras, Nour Elhouda
Résumé: Avec le développement des technologies de l’information et des échanges électroniques, en outre, le principal problème reste au niveau des échanges de données sur Internet, la garantie de confidentialité des informations est devenue très faible. Dans ce contexte, de nombreuses solutions informatiques sont apparues, mais elles sont encore insuffisantes, ce qui a conduit à l’émergence de la stéganographie de l’information comme solution complémentaire afin de contribuer à la sécurité des images partagées sur le réseau. Le travail présenté est reflété dans notre mémorandum sur la stéganographie dans images numérique, visant à renforcer la sécurité et la protection des droits d’auteur, et à assurer la confidentialité des communications entre deux parties. Nous avons abordé une méthode puissante pour tatouer des images numériques, qui est LSB, où la présence du message secret dans l’image stego est pratiquement imperceptible et indétectable. En particulier, il est nécessaire de s’assurer que la qualité de l’image ne se dégrade pas visuellement, nous voulons donc présenter une application pratique qui peut fournir des performances entre la capacité d’insertion, l’imperceptibilité et la robustesse et du stéganographie. Pour que le processus soit efficace, il doit être robuste, imperceptible et une statistique invisible.2023-01-01T00:00:00Z