Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Omar Gatera has a total of 74 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    FAST-TB: An FPGA-Accelerated Systems for AI-based Tuberculosis Detection on KV260

    by Arthur Nathaniel Mwang’ONda, Joel Mandebi Mbongue, Omar Gatera, Desire Ngabo

    Tuberculosis (TB) is a significant global health challenge, claiming over 1.5 million lives annually, particularly in regions with limited access to healthcare. Field-Programmable Gate Arrays (FPGAs) provide flexible and efficient hardware acceleration for a wide range of applications, including Deep Learning. In this paper, we present an FPGA-Accelerated Systems for AI-based Tuberculosis Detection on Kria KV260 (FAST-TB). It is a System-on-Chip (SoC) architecture that leverages FPGA and Deep Learning techniques to classify chest X-ray images for TB detection. We develop a Deep Learning model that effectively distinguishes between TB-infected and non-TB-infected chest X-rays, achieving an impressive 99.5% average inference accuracy. The model is prototyped on a Kria KV260 SOM from Advanced Micro Devices (AMD). The prototyped solution runs image analysis with a throughput up to 1864.39 frames per second, power consumption of 5.598 watts, and latency in the order of 0.25457 seconds. The implementation on the AMD Kria KV260 SOM demonstrates the practical viability of this approach, highlighting its potential to improve TB diagnosis.

  • Open Access

    An edge AIoT framework for optimizing the placement of integrated soil sensors in agricultural wireless sensor networks

    by Egas José Armando, Damien Hanyurwimfura, Omar Gatera, Kwang Soo Kim, Athanase Nduwumuremyi

  • Open Access

    Performance Analysis of Machine Learning Techniques in Predicting Maize Crop Yield: Case Study of Kayonza District—Rwanda

    by Bobo Mafrebo Lionel, Richard Musabe, Omar Gatera, Celestin Twizere

    Climate change presents significant challenges to agriculture worldwide, leading to food insecurity and impacting rural livelihoods. Maize farming is especially vulnerable to extreme weather, such as heavy rainfall, high temperatures, soil acidity, humidity, and poor irrigation, which reduce crop yields and raise concerns about food security. The study aimed to develop a reliable and accurate machine learning method to predict maize crop yields using historical climate data to facilitate decision-making. This allows farmers and agronomists to forecast maize production based on past data for adaptation. A dataset from Meteo Rwanda and maize yield data from the Kayonza district, Rwanda, were used for training and testing. The weather data included annual mean temperature, maximum temperature, minimum temperature, rainfall, and soil temperature over the past thirteen years. The data were analyzed using machine learning techniques such as Random Forest regressor, Extreme Boost regressor, Gradient, Support Vector Machine, and LASSO (Least Absolute Shrinkage and Selection Operator). The results show that developing a high-yield crop depends on predicting and integrating climate variables, especially temperature and rainfall. Overall, Random Forest, Support Vector Machine, and Extreme Boost outperformed LASSO, with R2 values of 0.957, 0.955, and 0.953, compared to 0.256 for LASSO.

  • Open Access

    Performance Analysis of Machine Learning Techniques in Predicting Maize Crop Yield: Case Study of Kayonza District—Rwanda

    by Bobo Mafrebo Lionel, Richard Musabe, Omar Gatera, Twizere Celestin

    Climate change poses significant challenges to agricultural practices worldwide, affecting crop yields, food insecurity, and rural livelihoods. Maize crop farming is particularly vulnerable due to extreme weather conditions such as high rainfall, high temperature, soil acidity, humidity and unproper irrigation affecting crop yield and consider to be a source of hunger and food security concern. The aim of the study was to propose a reliable and accurate machine learning techniques to be used in the prediction of maize crop yield using historical climate and soil data for informed planning. This enables farmers, agronomists and decision makers to forecast maize crop yield based on historical data for adaptation. To come up with a comprehensive prediction model, historical dataset from Meteo Rwanda and Maize crop yield from Kayonza district-Rwanda were used in the training and testing. Weather data considered in this study were annual mean temperature, annual maximum temperature, annual minimum temperature, annual rainfall, soil temperature for the past thirteen years. The data collected were analyzed using Random Forest regressor, Extreme Boost regressor Gradient, support vector machine and least absolute shrinkage and least absolute shrinkage and selection ( LASSO) machine learning techniques. The results shows that random forest perform better compared to other models with an accuracy of R² 0.957, support vector machine 0.957, XGBoost regressor 0.953, LASSO 0.256 and can be recommended for prediction of maize crop yield. The random forest regressor will be adapted in design and development of prototype to improve farming decision making.

  • Open Access

    Optimizing LoRaWAN Performance Through Learning Automata-Based Channel Selection

    by Luka Aime Atadet, Richard Musabe, Eric Hitimana, Omar Gatera

    The rising demand for long-range, low-power wireless communication in applications such as monitoring, smart metering, and wide-area sensor networks has emphasized the critical need for efficient spectrum utilization in LoRaWAN (Long Range Wide Area Network). In response to this challenge, this paper proposes a novel channel selection framework based on Hierarchical Discrete Pursuit Learning Automata (HDPA), aimed at enhancing the adaptability and reliability of LoRaWAN operations in dynamic and interference-prone environments. HDPA leverages a tree-structure reinforcement learning model to monitor and respond to transmission success in real-time, dynamically updating channel probabilities based on environmental feedback. Simulation results conducted in MATLAB R2023b demonstrate that HDPA significantly outperforms conventional algorithms such as Hierarchical Continuous Pursuit Automata (HCPA) in terms of convergence speed, selection accuracy, and throughput performance. Specifically, HDPA achieved 98.78% accuracy with a mean convergence of 6279 iterations, compared to HCPA’s 93.89% accuracy and 6778 iterations in an eight-channel setup. Unlike the Tug-of-War-based Multi-Armed Bandit strategy, which emphasizes fairness in real-world heterogeneous networks, HDPA offers a computationally lightweight and highly adaptive solution tailored to LoRaWAN’s stochastic channel dynamics. These results position HDPA as a promising framework for improving reliability and spectrum utilization in future IoT deployments.


  • Open Access

    Internet of Things-Based Electromagnetic Compatibility Monitoring (IEMCM) Architecture for Biomedical Devices

    by Chiedza Hwata, Gerard Rushingabigwi, Omar Gatera, Didacienne Mukalinyigira, Celestin Twizere, Bolaji N. Thomas, Diego H. Peluffo-Ord’onez

    Electromagnetic compatibility is the capability of electrical and electronic equipment to function properly around devices radiating electromagnetic energy, without mutual disturbance. Hospital environments contain numerous devices operating simultaneously and sharing resources. Undetected electromagnetic interference can cause medical devices’ malfunctions, exposing patients and staff. Traditional monitoring is time-consuming and relies on expert interpretation. An Internet of Things-enabled embedded system architecture for remote and real-time monitoring of electromagnetic fields from medical devices is proposed. It integrates frequency probes, a Raspberry Pi 4, and a communication module. A three-month study conducted at Muhima District Hospital, Kigali, Rwanda, demonstrated the system’s effectiveness in monitoring electromagnetic field levels and cloud transmission. The signals were benchmarked against International Electrotechnical Commission and Rwanda Standards Board standards. Alerts are triggered when thresholds are exceeded, with results plotted on website and mobile interfaces. Emissions were highest at noon when the equipment was most active and lower after 1:30 PM, indicating reduced activity. The sample recorded statistics of electric fields include mean (1.0028), minimum (0.7228), and maximum (1.3515). Among the five filters evaluated, the Savitzky–Golay performed better, with MSE (0.235) and SNR (9.308). A 412 ms average latency and 24 h operation was achieved, offering a portable solution for hospital safety and equipment optimization.


  • Open Access

    Advancing rail safety: An onboard measurement system of rolling stock wheel flange wear based on dynamic machine learning algorithms

    by Celestin Nkundineza, James Ndodana Njaji, Samrawit Abubeker, Omar Gatera, Damien Hanyurwimfura

    Rail and wheel interaction functionality is pivotal to the railway system safety, requiring accurate measurement systems for optimal safety monitoring operation. This paper introduces an innovative onboard measurement system for monitoring wheel flange wear depth, utilizing displacement and temperature sensors. Laboratory experiments are conducted to emulate wheel flange wear depth and surrounding temperature fluctuations in different periods of time. Employing collected data, the training of machine learning algorithms that are based on regression models, is dynamically automated. Further experimentation results, using standards procedures, validate the system's efficacy. To enhance accuracy, an infinite impulse response filter (IIR) that mitigates vehicle dynamics and sensor noise is designed. Filter parameters were computed based on specifications derived from a Fast Fourier Transform analysis of locomotive simulations and emulation experiments data. The results show that the dynamic machine learning algorithm effectively counter sensor nonlinear response to temperature effects, achieving an accuracy of 96.5 %, with a minimal runtime. The real-time noise reduction via IIR filter enhances the accuracy up to 98.2 %. Integrated with railway communication embedded systems such as Internet of Things devices, this advanced monitoring system offers unparalleled real-time insights into wheel flange wear and track irregular conditions that cause it, ensuring heightened safety and efficiency in railway systems operations.

  • Open Access

    AI-IoT driven system for agricultural pest outbreak risk prediction

    by Jean Pierre Nyakuri, Celestin Nkundineza, Omar Gatera, Kizito Nkurikiyeyezu, Gervais Mwitende

    Invasive pests pose significant threat to agricultural production, specifically, Maize crops production, with severe implications for food security in many regions. Therefore, timely detection of pest development stages and accurate prediction of potential outbreak risks are essential for effective pest management. This study introduces a hybrid model, integrating Explainable Artificial Intelligence (XAI), a lightweight Convolutional Neural Network (CNN), and Fuzzy Logic (FL) for Fall Armyworm (FAW) pest detection and weather-based outbreak risk prediction. The model leverages a Lightweight CNN model “Tiny-MobileNet-SE” for image classification, XAI model based on Grad-CAM to provide transparency and interpretability of predicted image, enabling users to understand the decision-making process, as well as FL inference with environmental parameters including Temperature, Humidity, and Rainfall to predict FAW pest outbreak risks. The Tiny-MobileNet-SE model achieved impressive results of 98.6% accuracy, 98.5% F1-score, 98.6% Recall, 0.72 MB size, and 80 ms when deployed on Raspberry pi 5, outperforming state- of-the-art lightweight models including EfficientNetB0, Squeezenet, MobileNet_v2, MobileNet_v3, and ShuffleNet tested on the same settings, making it suitable for edge deployment. The proposed system offers a power efficient, scalable, and user-friendly solution for precision agriculture, providing actionable insights for pest management and contributing to sustainable crop protection strategies.


  • Top Cited Papers
    Open Access

    AI and IoT-powered edge device optimized for crop pest and disease detection

    by Jean Pierre Nyakuri, Celestin Nkundineza, Omar Gatera, Kizito Nkurikiyeyezu, Gervais Mwitende

    Climate change exacerbates the challenges of maintaining crop health by influencing invasive pest and disease infestations, especially for cereal crops, leading to enormous yield losses. Consequently, innovative solutions are needed to monitor crop health from early development stages through harvesting. While various technologies, such as the Internet of Things (IoT), machine learning (ML), and artificial intelligence (AI), have been used, portable, cost-effective, and energy-efficient solutions suitable for resource-constrained environments such as edge applications in agriculture are needed. This study presents the development of a portable smart IoT device that integrates a lightweight convolutional neural network (CNN), called Tiny-LiteNet, optimized for edge applications with built-in support of model explainability. The system consists of a high-definition camera for real-time plant image acquisition, a Raspberry-Pi 5 integrated with the Tiny-LiteNet model for edge processing, and a GSM/GPRS module for cloud communication. The experimental results demonstrated that Tiny-LiteNet achieved up to 98.6% accuracy, 98.4% F1-score, 98.2% Recall, 80 ms inference time, while maintaining a compact model size of 1.2 MB with 1.48 million parameters, outperforming traditional CNN architectures such as VGGNet-16, Inception, ResNet50, DenseNet121, MobileNetv2, and EfficientNetB0 in terms of efficiency and suitability for edge computing. Additionally, the low power consumption and user-friendly design of this smart device make it a practical tool for farmers, enabling real-time pest and disease detection, promoting sustainable agriculture, and enhancing food security.