A.I. for detecting urban fires at street level

The hackathon theme is “AI for Good” and will focus on the challenge of improving situational awareness for incident commanders responding to fires by enabling drones to detect urban fires from a street level view. This information gathering capability can help keep responders safe from unnecessary danger, and provide real-time information to firefighters, the public and the community.

The specific problem examined in this open hack is the challenge of identifying a fire within a panorama of an urban landscape. The drone camera has a limited field of view (FoV) so it cannot see the entire landscape in one take. This hack involves the dual challenges of navigating the drone around the landscape efficiently, and identifying regions as indicative of fire. 

For the purposes of the hack, the drone will be limited to two degrees of freedom as it processes its environment. As the following illustration shows, the drown can move up and down, as well as side to side, but cannot move closer or farther from the scene. The illustration also highlights field of view that the drone might actually have against the scene.

A FireDrone and its field of view

The hack will be conducted entirely within a simulated environment, where participants will be provided with software to control the movement of the drone and it’s camera as well as unlabelled scenes, accessible only thru the drone’s field of view.

The hack itself includes an API that enables you to directly leverage your solution with compatible drones, using their onboard camera. 



June 5, 2019: Hack available for registration. Teams can begin work on their prototypes.

June 12, 2019: Optional meet and greet at the Azure + AI conference in Orlando, Florida. Share your progress, interact with support team, ask questions and get inspired!

July 31, 2019: Online hackathon officially ends. All entries must be received by 11:59 PM PDT on this day.

August 5, 2019: Hackathon winners announced.

View full rules


  • Participants: Individuals (over 18 years in age); Teams; 
  • Countries: United States


  • All solutions must use Microsoft Azure Machine Learning or Microsoft Azure Cognitive Services as part of the solution.
  • All solutions must adhere to the Theme of the hack.
  • Microsoft employees, vendors and MVPs cannot actively participate but are encouraged to mentor teams.
  • Participants should be familiar with Cloud platforms like Azure, but this is not a requirement, however learning curves should be taken into account.
  • Any operating system (e.g. Linux or Windows), development language, development framework, development tools or hardware can be used.
  • Prototypes that stretch multiple cloud platforms will be accepted, however all machine learning or cognitive technologies used by the solution must be Azure based or hosted on Azure. It is recommended that the use of Azure Databricks and/or Azure Machine Learning service in Azure is considered for your prototype.
  • Other 3rd party machine learning tools hosted on Azure are welcome (e.g. commercial or open source tools) can be used to enhance the solution, however using 3rd party cloud based AI solutions will not be allowed.
  • The usage of the Azure cloud platform is encouraged, including other Microsoft cloud platforms like Microsoft 365.
  • Investment in your prototype is recommended but Microsoft will not be responsible for any costs or investments incurred by the teams. E.g. cloud consumption, electronics, etc.

In the spirit of the open hack, we require that all artifacts be made available on GitHub.

Your participation may be required at future Microsoft marketing events where we may select your project to be showcased.


Ciprian Jichici

Ciprian Jichici

Zoiner Tejada

Zoiner Tejada

Judging Criteria

  • Vision and Clarity
    How clearly the project is articulated.
  • Creativity and Innovation
    How unique is the solution. Are you enhancing something familiar with AI or is this an entirely new approach? How unexpected is the project’s solution to the problem?
  • Demo Functional
    The demo works, presented well and was useable by judges
  • Performance
    performance metrics of solution (e.g., internal metrics on computer vision, recognition of map objects)
  • Potential Real-World Impact
    Is it just for fun, or does it have a real future How easy is it to scale the project into a real-world solution? What needs to be refined for a production quality version?
  • Use of Azure Platform
    Machine learning or cognitive technologies used by the solution must be Azure based or hosted on Azure. It is recommended that the use of Data Bricks and/or Machine Learning Services in Azure is considered for your prototype.