In a mind-blowing PhD public defense held at Nkumba University on Friday, April 11, 2025, Lillian Tamale, a PhD Candidate in Computing, unveiled a pioneering deep learning model that harnesses computer vision and image classification techniques for the early detection of aflatoxins in ground nuts. By enabling real-time diagnosis through a mobile-based application, Tamale’s work interwinds cutting-edge artificial intelligence with accessible agricultural tools, potentially reducing aflatoxin-related health risks and post-harvest losses in rural communities.
Titled “Artificial Intelligence Agritechnology and Post-Harvest in Uganda: A Case of Aflatoxin Early Detection in Groundnuts among Small-Scale Farmers of Teso Region,” her dissertation captivated the examination panel, faculty, and fellow researchers with its real-world impact and technological depth.
During her presentation, Tamale detailed how she leveraged computer vision algorithms and trained a deep learning convolutional neural network (CNN) model using a class-structured image dataset to distinguish between healthy, moldy, pest-infested, and discolored groundnuts. The model applied advanced feature extraction techniques to classify contamination levels with remarkable accuracy.
“This section presents the descriptive statistics on post-harvest handling practices, followed by the output of the deep learning classification model, which applies advanced feature extraction techniques in a computer vision framework to accurately identify aflatoxin-related contamination in groundnut images,” she explained, while illustrating the model’s architecture and performance metrics.
Panelists raised critical questions about environmental factors, such as seasonal moisture levels, which influence the appearance and quality of ground nuts. One examiner commented, “In rainy seasons, groundnuts often get mixed with soil. Can your model still differentiate infections under such conditions?”
Tamale confidently responded, “Yes. We ensured that the dataset included diverse images representing real-world scenarios. The model was trained using images across various stages and conditions. The emphasis was on post-harvest quality, where visual symptoms of aflatoxin contamination become more detectable.”
She further elaborated, “I hypothesized the interaction between environmental and biological data features, applying feature extraction and classification to evaluate contamination. This allows for real-time detection using a mobile-based computer vision system.”
Her deep learning model, she noted, demonstrated high classification accuracy and the potential to be deployed via a mobile application, enabling smallholder farmers to scan groundnuts using their smartphones and receive immediate diagnostic feedback.
“We’re living in the age of AI,” Tamale said. “Most farmers now own mobile phones. With this app, they will be empowered to make faster and safer post-harvest decisions.”
The highlight of the defense came when Prof. Jude Lubega, Vice Chancellor of Nkumba University, confirmed her success: “You have passed,” he declared. “You now join the distinguished rank of PhDs. Congratulations. Please work closely with your supervisors to finalize the corrections within a month.”
In her closing remarks, an emotional Tamale revealed the personal and public health motivation behind her work: “Aflatoxins are deadly. They cause liver cancer and contribute to over 3,700 cases annually in Uganda,” she stated. “This model doesn’t only work on groundnuts—it can be adapted to other grains like cassava and maize, offering a scalable solution for the region.”
Her research focused on Uganda’s Teso sub-region, specifically Soroti, Serere, and Kaberamaido districts, where groundnut farming is a livelihood for many. “Current aflatoxin testing methods are lab-based, expensive, and inaccessible to small-scale farmers. My AI-powered solution is cost-effective, fast, and easy to use,” she emphasized.
Tamale issued a call for implementation support: “I’m appealing to partners who can help turn this model into a fully developed mobile application free and accessible to farmers across the country.”
She expressed gratitude to her supporters: “I was fully funded by RUFORUM. I sincerely thank my supervisors, Dr. Denis Ssebuggwawo and Dr. Drake Patrick Mirembe. I am also deeply grateful to Dr. David Kalule Okello, a renowned aflatoxin expert and Head of the National Groundnut Improvement Programme at the National Semi-Arid Resources Research Institute (NaSARRI), under the Uganda National Agricultural Research Organisation (NARO). Their guidance and expert insights significantly enriched my research, particularly in bridging my computing background with agricultural science. I also sincerely appreciate groundnut farmers from Soroti, Serere, and Kaberamaido districts, their invaluable field support played a critical role in data collection and the validation of my deep learning model.”
“To the Vice Chancellor, the examination panel, and everyone at Nkumba University, thank you for believing in me,” she concluded, receiving a heartfelt round of applause.
Expected to revolutionize crop production in Uganda, the research marks a significant advancement in AgriTech innovation, offering a transformative solution for post-harvest handling among small-scale farmers in Uganda.
Discussion about this post