Projects

This successfully addressed the growing problem of traffic congestion in urban areas in five main cities in the country, including Kampala. With the country's rapid urbanization, the project aimed to create an intelligent traffic management system to optimize traffic flow using real-time data and machine learning algorithms. Data from traffic cameras, sensors, and GPS devices was collected and processed through machine learning models, which analyzed traffic patterns and predicted congestion points. The system utilized reinforcement learning algorithms to adjust traffic signal timing based on current traffic conditions dynamically. This adaptive approach significantly reduced waiting times at intersections, minimized traffic jams, and improved overall traffic flow efficiency. The model was trained using historical traffic data from key intersections in Kampala, allowing it to make accurate predictions and continuously improve its decision-making over time. The system also integrated real-time feedback from local authorities and drivers, enabling manual intervention when necessary. This innovative traffic control system led to a measurable reduction in commuting time, fuel consumption, and vehicle emissions across Uganda's major urban centers. Leveraging data analytics and machine learning, the project provided a modern and scalable solution tailored to Uganda’s unique infrastructure challenges. The successful implementation of this system demonstrated the transformative power of artificial intelligence in urban mobility and laid the groundwork for the future development of smart cities in Uganda.

The project successfully created robust datasets aimed at improving the accuracy and speed of diagnosing crop pests and diseases. This project focused on compiling a large-scale, high-quality dataset using crop imagery and spectrometry data from various agricultural regions in Uganda. The dataset was designed to be the foundation for building machine learning models capable of detecting and diagnosing pest infestations and diseases in crops early. High-resolution images of crops captured under different lighting and environmental conditions were paired with hyperspectral and multispectral data obtained from advanced spectrometry devices. These datasets covered a wide range of crops, including maize, cassava, sweet potatoes, and bananas, which are staple foods in Uganda. Each image and spectrometric sample was meticulously labeled by experts, noting the type of pest or disease, the severity, and the stage of infection. The dataset development process involved field data collection, imagery preprocessing, and thorough annotation to ensure the accuracy of the training data. Data augmentation techniques were used to increase the dataset’s diversity, allowing machine learning models to generalize better across various environments and crop conditions. This dataset now serves as a critical resource for developing machine learning algorithms that can detect early signs of crop damage based on visual and spectral patterns. Thus,  through leveraging this dataset, future diagnostic tools will enable farmers and agricultural experts in Uganda to identify and manage crop diseases more efficiently, reducing crop losses and improving food security. This project laid the groundwork for next-generation precision agriculture tools that are vital for Uganda’s agricultural sustainability.

The project successfully tackled the pervasive issue of potholes on roads in Kampala, improving road maintenance and transportation safety. This project developed a system that employed deep learning algorithms to detect potholes in real-time using image data captured from cameras mounted on vehicles and drones. The system used a convolutional neural network (CNN) trained on a large dataset of road images labeled with pothole locations and characteristics. The model was designed to automatically detect potholes of varying sizes and depths under different weather and lighting conditions. By continuously learning from new data, the system improved its accuracy and efficiency in identifying potholes, even in challenging environments typical of Kampala’s road networks. The collected image data was processed and analyzed in real-time, allowing road authorities to map pothole locations and prioritize repair efforts based on severity. The system is also integrated with GPS data to provide precise coordinates of potholes, helping maintenance teams deploy resources more effectively. This project significantly enhanced the efficiency of road maintenance in Kampala by providing an automated, scalable, and cost-effective solution to pothole detection. The system reduced the need for manual inspections, accelerated repair schedules, and improved road safety for commuters. Utilizing deep learning and computer vision, the project demonstrated how advanced technology could address infrastructure challenges in urban areas, paving the way for smarter, more proactive maintenance strategies in Uganda.

This project successfully demonstrated the use of deep learning technologies to enhance COVID-19 diagnosis in resource-limited settings. The project focused on utilizing point-of-care ultrasound (POCUS), a portable and affordable imaging tool, to screen for lung abnormalities associated with COVID-19, with the assistance of deep learning models that automated the interpretation of ultrasound images. In this initiative, a convolutional neural network (CNN) was trained on a large dataset of ultrasound images from COVID-19 patients, capturing lung patterns indicative of infection, such as consolidation, pleural irregularities, and B-lines. The deep learning model was developed to identify these patterns with high accuracy, even when used by healthcare workers with limited ultrasound experience. The integration of AI into POCUS allowed for rapid, real-time analysis, making it an effective tool for screening patients in rural and urban healthcare settings in Kenya. The project was particularly impactful in resource-constrained areas, where access to advanced medical imaging such as CT scans and X-rays is limited. By providing a low-cost, portable solution for COVID-19 screening, the deep learning-assisted POCUS system facilitated faster diagnosis and triaging of patients, reducing the burden on healthcare facilities. This project contributed to improving Kenya’s COVID-19 response by offering a scalable, AI-powered solution that could be deployed in remote and underserved communities. The successful deployment of this technology demonstrated the potential for combining deep learning and point-of-care imaging to address public health challenges in low-resource settings, laying the foundation for future AI-assisted medical tools in Kenya’s healthcare system.

This project aimed to create a lightweight, portable ultrasound device equipped with advanced imaging capabilities and integrated smart technology to assist healthcare professionals in various clinical settings. The system utilized real-time imaging and sophisticated algorithms to provide precise anatomical visualization during procedures such as biopsies, catheter placements, and fluid drainage. By employing machine learning techniques, the ultrasound system was trained to recognize and highlight critical anatomical structures, improving the clinician's ability to navigate and perform interventions with greater confidence and safety. This feature was particularly beneficial in complex procedures where anatomical variations pose challenges. Field testing was conducted in multiple healthcare facilities, allowing practitioners to evaluate the system's usability and effectiveness in real-world scenarios. Feedback from healthcare professionals informed iterative design improvements, ensuring the system met the practical needs of users in different clinical environments. The Smart Portable Ultrasound System was designed for versatility, allowing it to be used across various medical specialties, including emergency medicine, radiology, and surgery. Its compact design and battery-operated functionality enabled healthcare providers to use the device in both urban hospitals and remote, resource-limited settings. Overall, this project demonstrated the potential of integrating smart technology with portable ultrasound systems to enhance patient care by facilitating minimally invasive procedures. Providing accurate, real-time guidance, the system not only improved procedural outcomes but also minimized patient discomfort and recovery times, paving the way for broader adoption of minimally invasive techniques in healthcare

This initiative sought to address the challenges of inefficient patient management, data fragmentation, and limited access to healthcare services, particularly in rural and underserved areas of Uganda. The system was designed to leverage machine learning algorithms to analyze patient data, predict health outcomes, and optimize resource allocation. By integrating data from various sources, including electronic health records (EHRs), laboratory results, and patient demographics, the system provided healthcare providers with actionable insights to improve decision-making and patient management. One of the key features of the system was its predictive analytics capabilities, which utilized historical patient data to identify at-risk patients and forecast potential health issues. This proactive approach enabled healthcare professionals to intervene early, improving patient outcomes and reducing the burden on healthcare facilities. The system also included a patient scheduling module that optimized appointment times and minimized wait times, enhancing overall patient satisfaction. Extensive field testing was conducted in collaboration with local healthcare providers to ensure that the system met the specific needs and workflows of Ugandan healthcare settings. Training sessions were held to familiarize healthcare workers with the platform, promoting user adoption and ensuring successful implementation. The successful development and deployment of this patient management system highlighted the potential of machine learning and data analytics in transforming healthcare delivery in Uganda. By improving patient management processes and facilitating better health outcomes, this project laid the groundwork for a more efficient, data-driven healthcare system that can adapt to the unique challenges faced by Ugandan communities.

The project project successfully implemented an innovative mobile application designed to facilitate cervical cancer screening and enhance access to healthcare services in Uganda. This initiative aimed to address the high incidence and mortality rates associated with cervical cancer by providing a platform for early detection and timely consultation with healthcare professionals. The mobile app integrated machine learning algorithms to analyze user-reported symptoms and health data, enabling the identification of individuals at higher risk for cervical cancer. By utilizing a user-friendly interface, the app guided users through a series of questionnaires and self-assessment tools to collect relevant health information. This data was then processed to generate risk profiles, which facilitated informed decision-making regarding further medical evaluation. Real-time tele-consultation features were embedded within the app, allowing users to connect with qualified healthcare providers via video calls or chat. This capability significantly improved access to specialist care, especially for women in remote areas where healthcare facilities may be scarce. Through enabling real-time consultations, healthcare providers could offer personalized guidance, discuss screening options, and address concerns related to cervical health. The project involved collaboration with local health authorities and stakeholders to ensure compliance with health regulations and community needs. Training programs were conducted for healthcare providers to familiarize them with the app’s functionalities and promote its use among patients. Overall, the successful development of this machine learning-enabled mobile app demonstrated the potential for technology to transform cervical cancer screening in Uganda. Enhancing accessibility to healthcare services and promoting early detection, the project contributed to improved health outcomes for women, ultimately aiming to reduce the burden of cervical cancer in the region.