http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/1196 2026-06-14T17:31:28Z 2026-06-14T17:31:28Z Nishat, Md Rifat Uddin Setu, Sadia Afrin http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17311 2026-06-14T03:59:09Z 2025-01-12T00:00:00Z

Snake Gourd Leaf Disease Detection Using Deep Learning Nishat, Md Rifat Uddin; Setu, Sadia Afrin This study explores a novel approach to identifying diseases in snake gourd leaves using advanced deep-learning techniques. The research focuses on five specific leaf conditions: Healthy, Powdery Mildew, Downy Mildew, Yellow, and Anthracnose. A custom dataset of leaf images, normalized to 224x224 pixels, forms the foundation of the study. Preprocessing techniques such as contrast stretching and gamma correction are employed to enhance image quality, ensuring robust inputs for the models. The study evaluates several cutting-edge deep learning architectures, including VGG19, MobileNetV2, and ResNet50V2, for classifying the leaf conditions. Among these, VGG19 emerges as the most promising model, achieving an impressive accuracy of 91.35%. This demonstrates the model’s potential for reliable disease detection in real-world applications. The proposed solution automates the disease detection process, offering a practical and scalable tool for early diagnosis in snake gourd cultivation. By enabling farmers to identify diseases at an early stage, this system helps prevent crop loss and improves agricultural productivity. The integration of artificial intelligence into precision agriculture, as demonstrated in this study, highlights its transformative potential in addressing challenges faced by modern farming. Furthermore, the research lays a solid foundation for future advancements in plant disease detection systems, offering insights into the development of more effective and accessible tools for agricultural applications. With its focus on leveraging state-of-the-art technology, this work contributes significantly to the growing field of AI-driven solutions in sustainable farming practices, ensuring better yields and enhanced food security. Project report

2025-01-12T00:00:00Z Das, Aprantar Mia, Md Hasan http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17307 2026-06-13T21:02:30Z 2025-01-12T00:00:00Z

A Deep Learning Framework for Precision Plant Nutrient Status on Different Crops Das, Aprantar; Mia, Md Hasan Plant nutrient deficiencies can significantly impact agricultural productivity and crop quality, posing challenges for farmers in identifying and addressing these issues early. This project focuses on deep learning techniques as well as computer vision to detect nutrient deficiencies in different crops. network.We utilized a convolutional neural network (CNN) model with transfer learning, specifically the VGG16 architecture, as the foundation of our approach. The pre-trained base layers of VGG16 were frozen during initial training to retain learned features, and custom classification layers were integrated for optimal performance. To enhance model accuracy and robustness, extensive preprocessing techniques were employed, including background removal, normalization, and data augmentation. A publicly available dataset from Kaggle served as the primary source for training and validating the model. Our experiments demonstrated high classification accuracy, providing actionable insights for identifying nutrient deficiencies in crops. The broader impact of this work lies in its potential to improve agricultural productivity and crop management. By enabling early and accurate detection of nutrient deficiencies, this solution empowers farmers to take timely corrective actions, ensuring optimal crop health and reducing the risk of significant yield losses. This project underscores the transformative potential of AI in agriculture, paving the way for smarter and more sustainable farming practices. Project report

2025-01-12T00:00:00Z Sultana, Sabbira http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17304 2026-06-13T21:02:31Z 2025-01-13T00:00:00Z

Blood Cancer Diagnosis using Deep Learning: Enhancing Accuracy in Leukemia Detection Sultana, Sabbira Blood problems are frequently discovered by visual inspection under a microscope. It might help classify a number of blood-related disorders in order to facilitate the identification of different leukemia conditions. Leukemia is the most common cancer that affects the body's red blood cells. Even in the absence of any outward signs, it can progressively damage all of the body's internal organs, leading to a host of other illnesses. Early blood disease identification is difficult since current technologies take longer. The creation of a technique that might aid in the classification of leukemia prediction is described in the article. The accurate identification of blood cancer is the primary objective of this study. Certain leukemia diseases, such acute myeloid leukemia (AML) and lymphoblastic leukemia (ALL), prevent cells from growing and protecting every component of the blood, which can cause cancer and other illnesses that can harm the blood in various ways. This study examined three distinct blood cell classifications: neutrophils, a subset of normal blood cells, and two distinct kinds of cancer cells. Clinical methods are not particularly good in predicting leukemia because symptoms include fatigue, sickness, fatigue, and loss of appetite. In order to obtain the best results for identifying blood cancer utilizing the most precise categories possible using a deep learning approach, the strategy was designed to predict blood cancer using DL techniques for picture automated identification and splitting methodology. used 6000 photos in this study. In the process, we showcased and compared several deep learning models, such as Inception V3, Mobile Net V2, InceptionResnetv2, and VGG16 & VGG19. The assessments that were done on the all five models showed that get the best accuracy of 100% on the dataset, which is excellent. Following accuracy in the detection of a certain kind of blood cancer. Project report

2025-01-13T00:00:00Z Faysal, Md Minhajul Abedin http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17302 2026-06-13T21:02:28Z 2025-01-13T00:00:00Z

An Innovative Smart Parking system Faysal, Md Minhajul Abedin Complete details of my concept, "An Innovative Smart Parking system," may be obtained in the accompanying results. This paper goes into great detail on the processes that helped transform the plan into a working website. Administrators, parking attendants, and automobile owners are the three modules into which people can be categorized inside the system. An intelligent parking system for reservation that offers consumers a straightforward online parking space reservation procedure is the notion that has been proposed. It eliminates the need to spend time looking for parking at places of business. In order to allow users to park their automobiles and contact the administration, this project offers a web-based booking system. The automobiles that are supposed to be parked in the lot are tracked by this initiative. An additional benefit provided by this system is the opportunity to cancel a car reservation. Users can cancel their book space at any time, according to the administrator. Additionally, users have the option to pay fees that the administrator has determined. After making the debit payment, users receive a parking out message along with their unique parking ID. Users and the administrator can quickly update their different data. Additionally, printed parking recites for the admin, user, and parking attendant modules are included. In the admin dashboard, you may add a parking attendant, search for a new parking spot, and change all the data. Additionally, users have the ability to submit parking requests to the parking attendant and administrator. Additionally, parking attendants can accept or reject requests. Our online application is developed in HTML for user interfaces, CSS for web page layout, and PHP with a reliable MYSQL database backend that uses XAMMP. All that is required for systems to establish up our system applications are desktop computers and online access; costly software and hardware elements are not required. If you have the right login information, you may utilize our infrastructure as a worldwide repository and as easily accessible apps from anywhere. With a typical internet-accessible network, almost every user may use our platform-independent solution at any time and from any location. And we could change our system to meet certain needs. Project report

2025-01-13T00:00:00Z