Ayodeji Falayi works as a researcher at Towson University. His research work primarily focuses on integrating Computer vision and Artificial Intelligence for Food safety. He hails from Ijan Ekiti in Ekiti State and earned a Bachelor of Technology degree in Food science and Technology (FST) from the Federal University of Technology Akure (FUTA) in 2016. He proceeded to the Katholieke University of Leuven (KU Leuven), Belgium for his First master’s degree in Bioscience Engineering while Majoring in Food technology. He later proceeded for another master’s degree in computer and information technology at Towson University, USA, majoring in Machine vision and AI. During his First Master thesis, Ayodeji worked on Physical and Biological factors that influence the development of mealiness in blueberry (Vaccinium corymbosum L.) He became interested in the connection between enzyme activity and mechanical properties of blueberries leading to mealiness. Ayodeji interest in roles of computer vision for object detection and increased efficiency in food system spurred from his first thesis where he wants faster detection of mealy blueberries rather than the time-consuming destructive method for juiciness determination which was investigated as preliminary indicator for mealiness. Ayodeji proceeded for his second master’s thesis where he focusses on Plant Disease Recognition Method Based on Fusion of Images and Graph Structure Text, He introduced a technique called feature deconstruction and recombination, which enables the model to learn picture features at various levels of detail in order to properly identify distinct test images. The disease feature words were extracted from the disease text description information, which consisted of continuous vectors. These words were then recombined into the disease graph structure text. The graph convolutional neural network (GCN) was subsequently used for feature learning. Ultimately, he developed a methodology for identifying vegetable diseases by combining photos and graph structure text. Ayodeji research continues with his PhD work on artificial intelligence in machine vision for postharvest detection of disorders in horticultural products. He applies Deep learning algorithms with RGB datas for Quality detection of fruit and vegetables. Excerpts:

From Ekiti to USA, it has indeed been a journey filled with various experiences. Tell us about it.

First, thank you very much for this opportunity. I was born and raised in the ancient town of Ikole Ekiti, in Ekiti State where I started my primary school education at Oluwanifesimi Nursery and Primary schools, Ilupeju Ekiti. I proceeded to St.Theresas catholic school, Ikole Ekiti for my secondary education. In 2011, I gained admission to the prestigious Federal University of Technology, Akure (FUTA) to study Food science and Technology. I graduated in 2016 and undertook the compulsory one-year NYSC programme which I completed in 2017 in Ado-Ekiti, Ekiti State. I have always had during my Undergraduate program to have my graduate education abroad. Prior to my graduation, I have been making plans on my post graduate studies in the Joint master’s program of Food technology (IUPFOOD) at Ghent / Katholieke University of Leuven, Belgium. My interest spurred from Couple of my Exceptional Undergraduate Professors who had their Postgraduate studies in the same institution. Ghent/ Leuven master’s in food technology ranks

the second best in Europe and 5^th in the world. This solidifies my interest in the program and was determined to go for it. Based on my experience, Belgium is an amazing place with a lot of opportunities and tuition fees are relatively cheap. So, I jumped at the opportunity and has been a positive one for my career path.

With your continued research on diverse approaches computer vision and AI driven food system, how do you think this can improve food safety and quality assurance? what do you hope for?

AI has the potential to optimize different aspects of the food system, such as production and distribution, by minimizing waste, improving efficiency, and strengthening food safety and security. Annually, foodborne infections result in 600 million instances and contribute to 420,000 fatalities, along with a staggering loss of 33 million years of life when accounting for disability. Despite extensive efforts, we have not been successful in significantly reducing the prevalence of these diseases. Infection rates remain stable in numerous nations, including the United States, which has failed to achieve its targets for reduction. Combatting foodborne infections poses challenges due to inadequate surveillance systems in many regions, the extensive diversity within our food supply chain, and the presence of several pathogens with distinct modes of transmission and contamination.

Continuous research and innovations are ongoing on food system. (UC Davis) recently employed YOLOv4 for the purpose of pathogen identification. Research has demonstrated that the combination of AI and optical imaging can rapidly and precisely detect hazardous bacteria in food. This development has significantly transformed the process by making it more straightforward, cost-effective, and time-efficient compared to traditional methods of pathogen detection. Using digital photographs of romaine lettuce, the AI software demonstrated exceptional accuracy (specifically, 94% precision) in detecting minuscule clusters of bacteria. This method proved to be far faster than other approaches in swiftly identifying E. coli contamination. Food producers now have an efficient new tool to rapidly detect and address contamination, significantly reducing the lengthy waiting periods often required by previous approaches that cultivate microbial cultures.

In one of your projects, you worked on a defect grading method based on deep learning to identify the number and area of defects in an apple image. You used the BiSeNet V2 network to build a defect detection and segmentation model, and also used the YOLO V4 network to correct the results of BiSeNet V2 detection, Tell us more about this.

Yes, in my PhD work, I am currently involved in the process of identifying and classifying high- quality horticulture goods in real-time. At present, a belt or rollers are usually used to carry apples in the apple grading method. Usually, this causes financial losses due to damaged or overpriced fruits that are either too soft or otherwise compromised. The development of specific fruit trays

allowed for the real-time detection and classification of high-quality apples while also protecting them from bruising during image capture. To successfully isolate the faulty areas of imperfect apples, a new method was presented that makes use of the BiSeNet V2 deep learning network. The BiSeNet network could basically balance the relationship between real-time performance and accuracy. At 99.38%, BiSeNet V2 outperformed DAnet and Unet by 0.15% and 0.18%, respectively, in terms of Mean Pixel Accuracy (MPA). A model pruning method was used to optimize the structure of the YOLO V4 network. When it came to pinpointing where apple defects were located in images, the pruned YOLO V4 network was much more accurate. A method for accurately calculating the defect area was subsequently developed, which involved comparing images of the apple’s surface with the actual fault area. Separate fruit trays were eventually arranged according to the number and magnitude of flaws visible in the images. The results showed that with an F1 score of 94.76, the average accuracy for classifying apples was 91.34%. With this, a wide variety of uses could be found for commercial fruit tray sorting and grading devices.

In another project of yours, you implemented a SVM (Support vector machine) and YOLO for a mobile food grading system, tell us more about how this AI models have been useful in Food system.

The quality and safety of food are significant concerns for society, as they are crucial not only for individual health but also for societal development and stability. Ensuring the quality and safety of food is a multifaceted procedure. All steps of food processing, including cultivation, harvesting, storage, preparation, and consumption, should be considered. Grading is a crucial procedure for regulating the quality of food. As part of this study, I proposed a visual method for assessing food grading that could be easily accessed using mobile devices. The suggested method achieves this by capturing images of bananas as they move along conveyors. A banana grading system utilizes a two-tiered machine learning image processing system, with one layer operating on edge devices and the other on cloud servers. The initial layer is a Support Vector Machine (SVM) that is trained using a feature vector consisting of data on both texture and color. This SVM is employed to accurately categorize bananas. Next, we will utilize a You Only Look Once (YOLO) v3 model to detect the section of the peel that is not in ideal condition and distinguish between partially ripe and totally ripe inputs. The trial findings indicate a total accuracy of 96.4%, with the first layer achieving a performance of 98.2% and the second layer achieving 86.6%.

What are the main challenges you foresee in implementing Machine Vision and AI in the food system? Will A.I replace human jobs? How close are we?

As with any contentious topic, there will inevitably be individuals who support it and others who oppose it. Artificial Intelligence is indistinguishable. Indeed, with the introduction of new artificial intelligence techniques and the increasing coverage of news related to them, the division between the two groups will inevitably expand. According to many market research experts, AI has the capacity to generate various beneficial transformations in society, such as heightened efficiency, enhanced healthcare, and expanded educational opportunities. However, immediate adaptation is necessary at this moment. Some individuals, particularly those employed in labor-intensive occupations, argue that AI and robotics are a disruptive influence that mostly results in job displacement. However, robots and AI technologies have the potential to generate several novel occupations and effectively address intricate issues, hence enhancing the convenience and ease of our daily lives. The verdict is still pending, but the prevailing inclination is towards AI being a beneficial force rather than a detrimental one.

Although Africans may face bad perceptions on the global stage, individuals like yourself are actively enhancing the country’s reputation. How does this positive recognition make you feel, and how significant is it to alter the unfavorable narrative?

It is concerning how we are often subjected to stereotypes without people taking the time to understand who we truly are. During my initial years in Belgium, I encountered some instances of stereotyping. However, I promptly created my own identity and carefully selected my social interactions and destinations to prevent any potential embarrassment. I crafted my own narrative and exerted a certain level of influence over your perception of me. Fortunately, I am surrounded by highly educated individuals who are respectful, intelligent, well-traveled, and experienced. This, to a significant degree, protects me from the typical commotion and activity on the streets. I believe that each of us has a responsibility to consistently exhibit our finest behavior as individuals. As Africans, it is our responsibility to assume specific positions in order to have control over the narratives we convey and the message we present to the outside community. By engaging in such actions, we can influence the international society to embrace our perspectives and desired image.

What are your plans after graduation and how will they be useful to both Nigeria and the United States?

I am yet to determine my future residence upon the completion of my Doctoral studies at Towson University. I am interested in furthering my research in Intelligent food systems. Although my main focus is on continuing my research in the USA, I also have plans to collaborate with academic and research institutions in Nigeria. Researching in Nigeria offers the opportunity to collaborate with exceptional students and tap into the abundant talent pool available in the country. One drawback is the absence of a robust scientific community, which necessitates collective efforts to establish one. Furthermore, Nigeria faces a scarcity of research funding.