Our inspiration for project was to increase the amount of safety in corporate buildings and ensure better search and rescue operations during emergency situations, primarily fires. Our project primarily focused on creating an artificially intelligent mapping system that takes into account the current environmental hazards of each room and outputs the most optimal path for victims to exit the building. Link:

What it does

Escape is a smart mapping system that provides the user with visual aids towards optimal routings out of the building and away from hazardous flames. Our hardware project accomplishes this by creating a smart lighting system delimitating the most optimal path out of a building based on current room conditions. Escape also incorporates noise extraction and voice classification to detect if there are any victims still trapped in certain rooms. We also decided to create a web application to visualize how our hardware project could scale given larger floor plans.

How we built it

We used Java to generate random floor plans and used optimized pathfinding algorithms to find the best exits out of a building given current room constraints. We also used the VOiCES dataset, Librosa, Keras, and Python to build and train our classification neural network to detect coughing or screams from help from noisy audio inputs. For our frontend, we used HTML, CSS, JQuery, and JavaScript to create the animations and and the UI. Finally, we used Node.js and Ajax to connect the backend to the frontend.

Challenges we ran into

We originally wanted to use the SnapDragon processing board but we had trouble connecting the board to wifi. We ended up switching over to a Raspberry Pi 3. This was also our first time building a hardware hack for all of our members, and we though the process was pretty challenging but rewarding. We had a lot of fun learning new techniques and libraries, such as noise extraction with Keras and maze generation algorithms.

Accomplishments that we're proud of

  • Built our first hardware hack!
  • Created a software web application that displayed a scaled version of our project
  • Learned different machine learning and AI techniques

What we learned

No sleep is not good, uwu

What's next for Escape

We would like further extend our data visualization aspect to incorporate 3D building floor plans as well as predicting the spread of the fire during our path finding algorithms.

Built With

Share this project: