June 7, 2020 – Prescila Aramouny and Michael Khoury, two computer science students at Notre Dame University-Louaize (NDU), were part of a team to win 3rd place in the first Beirut AI Hackathon. Aramouny and Khoury collaborated with two AUB students to form their team, Growbots – making them the only participating team whose members did not initially know each other and met through BeirutAI's slack channel.
Due to the ongoing COVID-19 pandemic, the hackathon was held entirely online, with participants communicating on Slack and Zoom.
The hackathon’s challenge was to develop an idea to support one of Lebanon’s main sectors using Artificial Intelligence (AI) as a solution. Growbots chose the agricultural sector and using autonomous drone to detect weeds and defective crops in fields.
The ideal for the fully functional build would feature Raspberry Pi computer mounted on a drone to analyze and process the images in real-time. The drone itself would be autonomous, scheduled to follow specific routes using GPS navigation and come in two models: a weed detection model, which could identify different types of weeds and pinpoint their location on a virtual map; and an apple defect detection model, which would identify defects in apples.
“Implementing a prototype and pitching our idea was no easy task given the fact that we only had three days to build an idea from scratch,” said Aramouny. “We had no prior experience in machine learning and computer vision, so we had to develop our plan and learn simultaneously.”
With only two days to develop the idea, the final prototype consisted of code for an AI model that could detect defects in apples. “Implementation isn't straight forward, but there exist some pre-trained models that can make the job much easier,” said Aramouny.
The team won free credits from Oracle Cloud, which they can use to develop their prototype into a more functional model.