As a group of 4 people who met each other for the first time, we saw this event as an inspiring opportunity to learn new technology and face challenges that we were wholly unfamiliar with. Although intuitive when combined, every feature of this project was a distant puzzle piece of our minds that has been collaboratively brought together to create the puzzle you see today over the past three days. Our inspiration was not solely based upon relying on the minimum viable product; we strived to work on any creative idea sitting in the corner of our minds, anticipating its time to shine. As a result of this incredible yet elusive strategy, we were able to bring this idea to action and customize countless features in the most innovative and enabling ways possible.


This project involves almost every technology we could possibly work with - and even not work with! Per the previous work experience of Laurance and Ian in the drone sector, both from a commercial and a developer standpoint, our project’s principal axis revolved around drones and their limitations. We improved and implemented features that previously seemed to be the limitations of drones. Gesture control and speech recognition were the main features created, designed to empower users with the ability to seamlessly control the drone. Due to the high threshold commonly found within controllers, many people struggle to control drones properly in tight areas. This can result in physical, mental, material, and environmental damages which are harmful to the development of humans. Laurence was handling all the events at the back end by using web sockets, implementing gesture controllers, and adding speech-to-text commands. As another aspect of the project, we tried to add value to the drone by designing 3D-printed payload mounts using SolidWorks and paying increased attention to detail. It was essential for our measurements to be as exact as possible to reduce errors when 3D printing. The servo motors mount onto the payload mount and deploy the payload by moving its shaft. This innovation allows the drone to drop packages, just as we initially calculated in our 11th-grade physics classes. As using drones for mailing purposes was not our first intention, our main idea continuously evolved around building something even more mind-blowing - innovation! We did not stop! :D

How We Built it?

The prototype started in small but working pieces. Every person was working on something related to their interests and strengths to let their imaginations bloom. Kevin was working on programming with the DJI Tello SDK to integrate the decisions made by the API into actual drone movements. The vital software integration to make the drone work was tested and stabilized by Kevin. Additionally, he iteratively worked on designing the mount to perfectly fit onto the drone and helped out with hardware issues. Ian was responsible for setting up the camera streaming. He set up the MONA Server and broadcast the drone through an RTSP protocol to obtain photos. We had to code an iterative python script that automatically takes a screenshot every few seconds. Moreover, he worked toward making the board static until it received a Bluetooth signal from the laptop. At the next step, it activated the Servo motor and pump. But how does the drone know what it knows? The drone is able to recognize fire with almost 97% accuracy through deep learning. Paniz was responsible for training the CNN model for image classification between non-fire and fire pictures. The model has been registered and ready for use to receive data from the drone to detect fire. Challenges we ran into: There were many challenges that we faced and had to find a way around them in order to make the features work together as a system. Our most significant challenge was the lack of cross-compatibility between software, libraries, modules, and networks. As an example, Kevin had to find an alternative path to connect the drone to the laptop since the UDP network protocol was unresponsive. Moreover, he had to investigate gesture integration with drones during this first prototype testing. On the other hand, Ian struggled to connect the different sensors to the drone due to their heavy weight. Moreover, the hardware compatibility called for deep analysis and research since the source of error was unresolved. Laurence was responsible for bringing all the pieces together and integrating them through each feature individually. He was successful not only through his technical proficiencies but also through continuous integration - another main challenge that he resolved. Moreover, the connection between gesture movement and drone movement due to responsiveness was another main challenge that he faced. Data collection was another main challenge our team faced due to an insufficient amount of proper datasets for fire. Inadequate library and software versions and the incompatibility of virtual and local environments led us to migrate the project from local completion to cloud servers.

Things we have learned:

Almost every one of us had to work with at least one new technology such as the DJI SDK, New Senos Modulos, and Python packages. This project helped us to earn new skills in a short amount of time with a maximized focus on productivity :D As we ran into different challenges, we learned from our mistakes and tried to eliminate repetitive mistakes as much as possible, one after another.

What is next for Fire Away?

Although we weren't able to fully develop all of our ideas here are some future adventures we have planned for Fire Away : Scrubbing Twitter for user entries indicating a potential nearby fire. Using Cohere APIs for fluent user speech recognition Further develop and improve the deep learning algorithm to handle of variety of natural disasters

Built With

Share this project: