http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/1196 2026-05-03T19:02:11Z 2026-05-03T19:02:11Z Anzum, Tanvir http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17122 2026-05-03T03:40:10Z 2025-05-14T00:00:00Z

A Comprehensive Survey on Chilli Leaf Disease Detection Techniques Using YOLO Anzum, Tanvir Chilli (Capsicum annuum) is an economically vital crop extensively cultivated in countries like Bangladesh, India, and Indonesia. However, the crop is highly susceptible to fungal, bacterial, and viral diseases that significantly affect yield and farmer income. Traditional disease detection methods are manual, slow, and error-prone, particularly in rural areas where expert agronomists are scarce. To address this issue, this research proposes a real-time chilli leaf disease detection system based on advanced object detection models from the YOLO (You Only Look Once) family. Three models—YOLOv8s, YOLOv9s, and YOLOv10s—were trained and evaluated on a custom-annotated chilli leaf dataset. The YOLOv10 (small) model achieved the highest overall performance, with a mAP50 of 96.9%, mAP50–95 of 91.2%, and the fastest inference speed of 7.3 milliseconds. Compared to YOLOv8s and YOLOv9s, YOLOv10s demonstrated superior accuracy and lower computational load (24.5 GFLOPs), making it the most suitable model for mobile deployment. The best-performing model was optimized using TensorFlow Lite and integrated into a Flutter-based Android mobile application, enabling offline disease detection directly from smartphones. This approach empowers farmers with immediate, on-field diagnostic capabilities without relying on network connectivity or high-end hardware. Additionally, it promotes sustainable agriculture by supporting targeted pesticide use, thereby reducing environmental impact. Through careful model selection, optimization, and mobile app development, the study successfully bridges the gap between research and real-world application, providing a scalable, efficient, and farmerfriendly solution for chilli disease management. Future improvements could focus on multi-crop support, explainable AI integration, and multilingual user interface enhancements to broaden accessibility and impact. Project Report

2025-05-14T00:00:00Z Tithi, Nosrat Jahan Haque, Kazi Zakiul http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17120 2026-05-03T03:39:36Z 2025-05-14T00:00:00Z

Mental Health Prediction Among Medical Students Using Machine Learning Tithi, Nosrat Jahan; Haque, Kazi Zakiul Mental health is a critical aspect of well-being, influencing individuals' ability to function in daily life and cope with stress. Medical students struggle with depression alongside their demanding academic schedule and long workdays since these circumstances harm their academic results and general health condition. The objective of this research considers the application of machine learning algorithms to detect depression severity levels in medical students. Data contains a total of fourteen features with demographic information from age and year of study and psychological indicators such as interest scores compared to pleasure levels in addition to ratings regarding fatigue and sleep problems as well as changes in appetite and concentration abilities and PHQ-9 scoring scales. The depression severity makes up the target variable which divides into four groups: Severe Depression, Moderate Depression, Mild Depression and Moderately Severe Depression. The assessment of this task incorporated numerous machine learning models which consisted of Gaussian Naive Bayes, Random Forest Classifier, AdaBoost Classifier, Logistic Regression and Support Vector Classifier (SVC). Random Forest Classifier delivered the highest accuracy of 99.04% while Logistic Regression reached an accuracy of 98.80% yet Gaussian Naive Bayes obtained 97.36% accuracy. The accuracy of Support Vector Classifier at 95.44% was lower than the other models which included AdaBoost and its weakest performance of 79.41%. Predictions of depression severity among medical students demonstrate optimal performance when using Random Forest as the model selection because of its strong predictive capabilities. The study reveals crucial roles which machine learning serves mental health prediction alongside giving researchers a way to recognize at-risk students for proper early interventions and individualized treatments. Additional optimization and more targeted feature engineering techniques seem necessary to raise AdaBoost's results since there exists room for increased model accuracy. Researchers show that artificial intelligence performs effectively for mental health detection because it helps identify different depression severity levels within medical students who face elevated stress and health challenges Project Report

2025-05-14T00:00:00Z Mridul, Md. Mostak Ahamed http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17118 2026-04-30T02:33:51Z 2025-05-14T00:00:00Z

An Online-Based Hotel Booking and Room Rental Web Application - TourStay Mridul, Md. Mostak Ahamed In the midst of the fast growth of technology, online hotel and room reservations are already a necessity among many travelers. The culmination of this academic year presents a Hotel Booking and Room Rental Web Application that soothes the booking process for the owner and for those looking for place to stay. Owners can easily manage their listings, pricing, and booking on the asset, without needing to spend big bucks on the systems or expensive technical training. In this manner, small hotel businesses are managed to conserve costs and increase the potential guests. This website makes it so easy for travelers -each hotel and room search by a single platform. Users need just a few clicks and be able to view price comparisons, review amenities such as Wi-Fi or parking and see live availability. In addition, travelers are able to make their payment on the bookings securely and conveniently when using the built-in secure payment options available on the platform. With smart search and filter tools, customers can find the best deals that match their needs quickly. This web application aims to improve the hospitality industry by making bookings easier, reducing costs for hotel owners, and saving time for travelers. Designed with a focus on user-friendly features, scalability, and strong performance, it meets the growing demands of modern travelers and hotel businesses, creating a smoother and more efficient booking experience. Project Report

2025-05-14T00:00:00Z Jone, Israk Hasan Alam, Badrul http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/17115 2026-04-28T02:24:04Z 2025-05-14T00:00:00Z

Bangla News Headline Classification and Sentiment Analysis using Bangla Bert Jone, Israk Hasan; Alam, Badrul As a core task of natural language processing, text classification is widely used in many fields. Reading newspapers in daily life is a very common practice, but before reading, everyone first check the newspaper headlines. The news headline is important as it is supposed to provide an efficient way to grasp the flavor of the article and acts as a key factor to determine readers’ attitudes toward the article. In natural life, people naturally classify news in terms of themes and emotions using only headline impressions. However, in the age of constant and unorganized inflowing of digital news, manually clustering news by category and sentiment is a cumbersome job, particularly for Bangla news in which scarcity of automatic tools is found. Filling up this void, in this paper, we propose a state-of-the-art deep learning-based than we developed A Dual headed Classification model which Bangla News Headline Classification and Sentiment Analysis using transformer models. An end-to-end classification model was proposed using the BanglaBERT model to categorise (e.g., politics, religion, sports, other), and predict the sentiment polarity (positive, negative or neutral) of Bangla news headlines. Results were based on a dataset of 5033 training samples and 516 testing samples. Experimental results showed superior performance (with training accuracy of 98.18% and testing accuracy of 84.38% for aspect classification, and training accuracy of 97.17% and testing accuracy of 73.26% for sentiment analysis). Though there was a little bit of over fitting since the data set was small, the results clearly show strong potential in using pre-trained Bangla specific transformers in automated headline classification task. In the future, we will work on enlarging the dataset, further integrating data augmentation, and issuing real-time web applications for practical use of the system. Project Report

2025-05-14T00:00:00Z