海角社区-led DeepFire Team advances to final round in XPrize Competition
May 28, 2025
Environmental Sciences Professor Supratik Mukhopadhyay
BATON ROUGE - The world came a little closer to eliminating destructive wildfires last month, as teams in the advanced to the final round. Among them? Environmental Sciences Professor Supratik Mukhopadhyay and his DeepFire team.
In XPrize Wildfire, teams from around the world compete to develop new technologies that can rapidly and accurately detect and respond to wildfires. The DeepFire team鈥檚 entry is an advanced artificial intelligence engine designed to detect fires in their nascent stages, based on visual data and other information.
Take a closer look at DeepFire's work
鈥淎s we learned in January, during the Southern California Wildfires, forest fires can have a disastrous impact on life, properties, environment, and economy. It poses an existential threat to humanity. Going to the final of the prestigious Wildfire XPRIZE competition鈥� and winning the milestone award means a lot to our team,鈥� said Mukhopadhyay, a professor in the Department of Environmental Sciences.
鈥淚t provides an opportunity for us to deploy advanced AI-based wildfire prediction and detection technology that we developed painstakingly at 海角社区 over the years in a real-world scenario. This is synergistic with 海角社区鈥檚 Scholarship First agenda and shows that 海角社区 builds teams that win. This project shows the impact of 海角社区 both in Louisiana and beyond,鈥� he continued.
Mukhopadhyay and his team not only progressed to the final round, they were named one of six Milestone Awardees, meaning they receive an additional $50,000 prize. Eleven total teams advanced.
DeepFire is consistently able to predict fires with over 90 percent accuracy, Mukhopadhyay said. In the semi-finals, it identified fires at locations in the United States, Canada and Australia, all provided XPrize. In the final round, which will take place in August of 2025, teams will be required to detect fires in New South Wales, Australia 鈥� the site of catastrophic bush fires that killed 33 people and destroyed an area the size of Turkey in 2019 and 2020.
DeepFire鈥檚 system combines prediction, detection and spread modeling systems, all working together - the detection system uses information on fire prone regions and weather combined with an analysis of visual data to detect smoke. It then cross-references its findings with a predictive map, to increase or decrease confidence in whether a fire is occurring.
Mukhopadhyay said that, for the next round, the team plans to more tightly integrate the spread model, which predicts where and how a fire will spread through an area, into the other models.
They are also refining the AI engine to reduce latency in detecting fires.
This figure from DeepFire illustrates the risk of wildfires in various regions during a specific time frame. Calgary city proper exhibited a lower risk of wildfires. In contrast, smaller towns such as Cessford and the Acadia Valley, located in the northeast corner of the figure, faced a high threat of wildfires. Additionally, there were pockets around Medicine Hat in the Southeast that were also of concern.