Inspiration
We were inspired by the recent wild fires that took place in California. There was one story in particular, where a homeowner turned on his sprinklers before they left and saved their home. We wanted our users to have some level of control when there is an unexpected fire. This includes being able to automatically detect fires without 24/7 surveillance and being able to control their smart home to do certain actions.
What it does
Our web app opens a camera and scans for a fire using a pre-trained model hosted in the backend. We take 10 second clips and run it through the model. We then give back the edited clips to the dashboard where the users will have access too and can decide their next plan of action.
How we built it
We used a YOLOv5 pretrained computer vision model to recognize fires stored on our backend. Our front end was built on react and used fastapi.
Challenges we ran into
We initially had challenges on our design model and how we would implement the camera. We initially sought for a real time detection system but soon realized the difficulties. We then used a more discretized stream for our project
Accomplishments that we're proud of
Allow for the user to have a dedicated camera while looking at a dashboard and being able to formulate a plan
What we learned
We learned how to quickly design, implement, and execute a web app that has impact on the real world based on current events.
What's next for Flame Finder
Flame Finder will continue to grow and hopefully support real time stream and IOT access.
Built With
- fastapi
- react
- yolov5
Log in or sign up for Devpost to join the conversation.