Ever since I caught interest in computer science, I was fascinated about creating applications with machine learning. However, I never got the opportunity to learn it. I always learned some aspects of machine learning online but learning it was very tedious. But, by entering the competition, I thought I will get the motivation that I lacked throughout the online courses I attempted to take. And I have to admit, I did enjoy it much more than the online courses.
What it does
This drone "application" goes through each frame in a scene to detect fire and smoke so firefighters can get to where they need to as fast as they can. Drones have the ability to change their altitude and therefore get the best possible view to where a fire is. Unlike firefighters, drones are very agile and can move over buildings with ease. So giving a drone the job to detect where the fire is, is a faster way to save lives.
How I built it
I used Custom Vision thats is part of the azure cognitive services. I wrote all my code on azure notebooks in python 3.6.
Challenges I ran into
The Custom Vision iteration I created also detects smoke. This was to help create an efficient algorithm to predict where a fire can be and then test if that area has fire or not. The algorithm would avoid checking unnecessary locations for fire. It would save drone flight time. However, creating it was very difficult so I implemented a brute force algorithm.
Accomplishments that I'm proud of
This was my first time using Azure and its applications. I am proud that I know how to use parts of azure and implement them in my own python projects.
What I learned
I learned how to access Azure's custom vision api for training and prediction.
What's next for EndGame
If I was given more time, I would have researched possible algorithms online and implemented them in my project. In addition I would have made GUI for users of the application to have a more interactive experience.