• catrut

Building Artificial Intelligence to Free the Wild Blueberry


I’d like to introduce Patrick Hennessy, a thinker, a problem-solver and a friend; Pat has been a privilege to work with over the past year and continuing into the 2019-2020 school year. Patrick has shown his resilience and leadership in the Cultiv8 community pushing through barriers when we all felt “stuck” while remaining a positive, dynamic team player. We’d like to share a little about his journey and how Patrick’s work with Artificial Intelligence is innovating in the Wild Blueberry Industry.

Patrick Hennessy in Wild Blueberry Field of Study


Patrick completed his Diploma of Mechanical Engineering here at Dal AC, followed by a Bachelor in Mechanical Engineering at the Dal Halifax Campus. Pat spent two summers working at Dal AC in the engineering department, “My time at Dal AC gave me insight into the importance of Innovation in Agriculture and gave me valuable experience working in this field”.

“As the world population grows, food security is becoming a growing issue. Innovation will help ensure that we can produce enough food for the entire population. In Nova Scotia, innovation in agriculture allows our farmers to compete on a global scale. Food at our grocery stores now comes from all over the world, including areas where food input costs are lower (i.e. cheap labour). Innovation allows our farmers to produce more food for their money, which keeps their selling prices low enough to compete with food from other markets.”

Although Pat’s formal educational background is in mechanical engineering, he’s had a natural affinity for computer science. Intrigued and inspired, Pat taught himself all the material for his first computer programming course before the first class. “My mechanical engineering degree provided me with a very robust and varied skill set, but computer science has always been more of a passion for me”. His combination of curiosity, initiative and academic skill set employed him to take on an exciting project leading him to begin his journey in his M.Sc. under the supervision of Dr. Travis Esau who leads the Advanced Mechanized Systems (AMS) research program at DalAC.

The Prototype

Patrick’s research uses Deep Learning (a form of Artificial Intelligence) to detect weeds such as fescue and sheep sorrel in wild blueberry fields. Deep learning works by showing a computer thousands of labelled pictures of an object. Through this process, the computer “learns” the visual features of the object and can identify it again in unlabelled pictures.

AI Trained Camera With Herbicide Spray Nozzle

“I plan to implement this technology on a commercial herbicide sprayer for spot-application purposes. Currently, commercial sprayers apply a uniform coating of herbicide across an entire field. In most fields, the weeds only cover patches of the field and not the entire area. By using deep learning to recognize weeds, the sprayer nozzles can be turned on when a weed is detected and turned off otherwise. This will result in less herbicide usage overall.

As well, the AMS research team plans to implement this technology into an app so farmers can use their phones to identify weed species’, diseases, and plant growth stages.”

AI Identified Invasive Species

Pat anticipates that the project will result in cost savings for wild blueberry growers. The price of wild blueberries has declined over the past decade, which has resulted in financial burden for the growers. Herbicides are a big input cost for growers, so reducing the volume needed will help them remain financially sustainable.

Invasive Species to be Targeted by AI Technology

Overcoming Obstacles

After taking some time off school after his undergrad, Pat found he had to catch up on current innovations to make himself more competitive and up to date in industry innovations. Deep learning in particular was a new concept he had to learn which lead him towards his solution in reducing excessive herbicide waste and saving costs for wild blueberry farmers. In addition, to input from his supervisory committee, Pat read papers, watched a lot of YouTube videos, and bought a textbooks in self-directed study. “Honestly, YouTube helped me out the most in the early stages. There were a lot of great videos breaking down the concepts into simple ideas, and tutorials on how to get started practicing deep learning.”

“As with any big project, the key was to break it down into smaller, achievable goals to push forward” he explains, “I started by getting [reading/watching] as much information as I could. I then started practicing deep learning on simpler images like handwritten digits. I just kept diving deeper from there. I had a lot of support from my supervisor and research team along the way”.

“Don’t try to do everything at once. Break down your problem into smaller, achievable goals. As you complete them, those small victories will help motivate you to keep moving forward”, Pat advises. I think that communication and interviewing skills were the biggest takeaway from Cultiv8 for me. It’s given me the skills to dig deeper in conversations and helped me better understand the problems of the people my research is going to help .”

Pat presented some of his research at the American Society of Agricultural and Biological Engineers (ASABE) annual international conference in Boston this past July. The conference was attended by researchers from around the world, who discussed ongoing innovations in the fields of agricultural and biological engineering. Pat earned an award from the ASABE given to outstanding presentations in the field of mechanized systems.