Inspiration

We got lost so many times inside MIT... And no one could help us :( No Google Maps, no Apple Maps, NO ONE. Since now, we always dreamed about the idea of a more precise navigation platform working inside buildings. And here it is. But that's not all: as traffic GPS usually do, we also want to avoid the big crowds that sometimes stand in corridors.

What it does

Using just the pdf of the floor plans, it builds a digital map and creates the data structures needed to find the shortest path between two points, considering walls, stairs and even elevators. Moreover, using fictional crowd data, it avoids big crowds so that it is safer and faster to walk inside buildings.

How we built it

Using k-means, we created nodes and clustered them using the elbow diminishing returns optimization. We obtained the hallways centers combining scikit-learn and filtering them applying k-means. Finally, we created the edges between nodes, simulated crowd hotspots and calculated the shortest path accordingly. Each wifi hotspot takes into account the number of devices connected to the internet to estimate the number of nearby people. This information allows us to weight some paths and penalize those with large nearby crowds. A path can be searched on a website powered by Flask, where the corresponding result is shown.

Challenges we ran into

At first, we didn't know which was the best approach to convert a pdf map to useful data. The maps we worked with are taken from the MIT intranet and we are not allowed to share them, so our web app cannot be published as it uses those maps... Furthermore, we had limited experience with Machine Learning and Computer Vision algorithms.

Accomplishments that we're proud of

We're proud of having developed a useful application that can be employed by many people and can be extended automatically to any building thanks to our map recognition algorithms. Also, using real data from sensors (wifi hotspots or any other similar devices) to detect crowds and penalize nearby paths.

What we learned

We learned more about Python, Flask, Computer Vision algorithms and Machine Learning. Also about frienship :)

What's next for SmartPaths

The next steps would be honing the Machine Learning part and using real data from sensors.

Built With

Share this project:

Updates