RUN Repository

Communities in RUN Repository

Select a community to browse its collections.

Now showing 1 - 5 of 11

Recent Submissions

A Comparative Study of Two Convolutional Neural Network Models for Detecting Rice Plant Diseases Using Online and Local Image Data
(2024) Ojewumi, Theresa O.
Rice is one of the most widely staple foods around the globe, however, rice fields are severely affected by diseases, which can disrupt global food security. Early and accurate detection of rice diseases is essential for the recovery of such rice plants. Manually identifying rice plant diseases can be tedious and error prone. Artificial intelligence (AI) driven models, such as Convolutional Neural Networks (CNN) have proven very successful in the identification or detection of various crop diseases. This study, therefore, presents a comparative study of the effectiveness of two popular CNN architectures; ResNet and AlexNet for the detection of rice plant disease. The data used to train the models include a combination of rice leaf images that were gathered locally from a rice field/farm in Ede, Osun State, Nigeria, and froman online repository. The dataset consists of 5200 images classified into four classes: Bacterial leaf blight, Brown spot, Blast, and Healthy, each containing 1300 images. The effectiveness of the two trained models was measured using classification performance metrics including Accuracy, Precision, Recall, and F1-Score. The finding from the study showed that The ResNet has a test accuracy of 95.25% as against 92.91% for the AlexNet. The ResNet had 0.93 precision, while AlexNet recorded a precision of 0.24. For recall, the ResNet model had 0.98 while the AlexNet model had 0.23 and for the f1-score, the ResNet model had 0.95 while the AlexNet model had 0.24. Generally, the ResNet model outperformed the AlexNet model in detecting rice plant diseases, most significantly, brown spot disease.
TOMATODETECT: A MOBILE APPLICATION FOR DETECTING TOMATO LEAF DISEASES BASED ON VGG-16 CONVNET
(International Conference on Smart, Secure and Sustainable Nation, 2022) Ojewumi, Theresa O.
Tomatoes are a staple in Nigerian food, appearing in a wide range of recipes and providing several nutritional benefits such as Vitamin C, potassium, and lycopene, which help prevent heart disease and cancer. Nigeria is currently Africa's second-largest producer of fresh tomatoes, accounting for 10.8% of the continent's total. However, several diseases that infect the tomato plant have made it an endangered crop that needs special attention to reduce the massive loss of farmland. Farmers and other agriculture specialists go through tedious and time-consuming processes in visually inspecting crops that they suspect to be affected by various diseases in the real world, which does not guarantee accurate recognition and classification of specific plant diseases. Therefore, this study developed a mobile application to detect nine tomato leaf diseases and healthy tomato leaves. The Keras deep learning framework was used to develop two pre trained VGG-16 Convolutional Neural Networks (CNN or ConvNet) models. The model trained on the augmented data outperformed the model trained without augmented data, with an accuracy of 96.51%. Consequently, this DL model was selected and deployed in a developed mobile application that can accurately detect specific diseases and classify healthy leaves in a real-world scenario in tomato leaves. The selected VGG-16 pre-trained model was deployed into a mobile application environment by first converting it into a TensorFlowLite (TFLite) model adaptable in an android mobile application. To develop the mobile application, the kotlin programming language was used to design the logic of collecting data from users and sending them through the backend for verification with the firebase database, which handles the application’s storage and authentication. With this mobile application in the hands of tomato farmers, the outbreak and spread of diseases in tomato leaves can be detected early and prevented from becoming uncontrollable and threatening food security.
ESTIMATING SEMANTIC SIMILARITY IN YORUBA SENTENCES USING PATH-BASED METRICS
(University of Pitesti Scientific Bulletin, Series: Electronics and Computers Science, 2024) Ojewumi, Theresa O.
Measuring semantic similarity among texts is an important task in many Natural Language Processing applications such as information retrieval, text summarization. However, there is dearth of work in the development of the tool for Yoruba Language, and this has therefore limited the advancement of Yoruba Language Engineering. This study addressed the gap by using a knowledge-based approach based on lexical resources. A total number of 434 nouns were collected from home-domain. The nouns were grouped into hypernym semantic classes. The classes were thereafter organized in hierarchy to form a taxonomy for Yoruba nouns and concepts. The model for the measurement of semantic similarity in Yoruba sentences was thereafter developed using path-based similarity measurement between the concepts represented in the taxonomy. Using the model, the system was implemented using python programming language. The developed system was evaluated using accuracy mean opinion score, and a score of 73.2% was achieved.
Development of a phishing detection network using Ensemble Machine Learning Methods
(LAUTECH Journal of Computing and Informatics (LAUJCI), 2024-04) Ojewumi, Theresa O.
Over the years, phishing has been a major problem and has caused different people to lose sensitive information, hence leading to loss of financial assets. Different machine learning algorithms have been used in the assessment of phishing in different aspects: websites, emails, texts amongst others. However, phishing attacks continue to increase frequency and sophistication despite the numerous attempts to combat it, there is therefore a need for improved detection mechanisms. This study therefore assessed four machine learning algorithms (Random Forest (RF), Logistic Regression (LR), Naive (NB) and Support Vector Machine (SVM)), built an ensemble model with them and developed a system using this model to detect phishing websites. A dataset obtained from Kaggle machine learning repository containing 549,347 records of websites was split into two, 70% to train the ensemble model and 30% to test the model. Two categories of features were selected: Lexical based features and Domain based features of the URL. The performance of the four algorithms were evaluated using accuracy, precision, recall and f1-score. The model was implemented with Python programming language in Jupyter Notebook and 97.42% accuracy was recorded. The results obtained showed that proposed model is comparable to existing models with accuracies of 96%, 98%, 72% and 97% for LR, SVM, RF and NB respectively. The model was used to develop a user-friendly system where users can paste URLs in order to check the safety of the address. The system however is limited to HTTP protocols and might not be equipped to handle short URLs.
An ANFIS-Based framework DDoS attack
(Information Systems Journal, University of the West of Scotland, 2016) Ojewumi, Theresa O.
Purpose: In this paper, a simulation environment is set-up to generate distributed denial of service data in Virtual Knowledge Community (VKC) in order to study the pattern of attack in multi-agent environment. Design/Methodology/Approach: A threshold approach is used to construct the profile of the traffic for the agents in the network, and to identify anomalies whenever traffic goes out of profile. An Adaptive Neuro-Fuzzy Inference System (ANFIS) is used in classifying attack type and to determine its location. Four ANFIS are trained and tested in this research work to provide attack detection and classification. Findings: An Adaptive Neuro-Fuzzy Inference System (ANFIS) is used to categorise the datasets into attack and normal traffic using some standardized threshold values. Research limitations:/Implications: The study is only limited to detecting distributed denial of service attack in multi-agent environment. Practical Implications: The experiment and the simulation are carried out in a computer laboratory using Java Agent Development Framework and Adaptive Neuro-Fuzzy Inference System. Originality: An adaptive security is implemented in this study for detecting distributed denial of service attack in an agent-based virtual knowledge community.