Prototype  The working prototype which was presented, and won “Best Open Source” at the Thornton Tomasetti AEC Hackathon West 2019.

Prototype The working prototype which was presented, and won “Best Open Source” at the Thornton Tomasetti AEC Hackathon West 2019.

STROLL (APP)

This project was developed at the 2019 Thornton Tomasetti AEC Hackathon in Seattle. It won “Best Open Source Project” in the competition along with the Javascript module which powers the application.

This is an app which helps people find the most nature-filled walks/paths to take throughout the day in order to stimulate creativity and boost mental health.

GitHub
Project GitHub (App)
Project GitHub (Computational Module)

Tools:
Frontend: React
Backend: Google Maps API, Yelp API, Stroll Module
Deployment: Heroku

Team: Adam Chernick, Nate Holland, Petr Mitev, Jason Wheeler, Brandon Yu
Presentation: Link

Concept The goal of the application is to find the most “nature-filled” walk one could take, given a set of input parameters regarding the walk (speed, distance, etc.).

stack

The architecture of the application was split into two main categories: the user interface and interaction, and the backend engine which computes the possible paths. For the interface, React was used along with a number of other utilities to create a mobile-first web application which would stay active and responsive during a user’s walk. For the computational module, a number of APIs were used to compute a “nature score” and build a graph for quick computation of route and route permutations.

Stack The frontend is handled by React, while a series of APIs power the graph math engine which computes the possible paths.

PATH FINDING

In order to compute the “most nature-filled” path a user could take, we leveraged a number of Open Source tools and datasets to help us create a weighted graph which would quickly compute paths and variations. To aggregate these APIs and domain-specific computation, we created the Stroll Javascript module. The module allowed us to start with a location, create a point grid around it for sampling, and then assign a “nature score” to each point and path within that grid. The platforms listed below were used to compute the “nature score”.

Yelp
The Yelp API was used to get a list of nearby parks and green spaces.

Google Maps
The Google Maps API was used to compute human-walkable routes, and to gather “Street View” images from the points in our point grid. The images were then analyzed for dominant colors (using node-vibrant) and dominant objects (using tensorflow.js and the pre-trained coco-ssd model with reinforcement learning). This allowed us to compute how much nature was roughly “visible” at a given point.

Path-finding  The  “Stroll” Javascript module  was created to provide a high-level interface into computing a “nature score” for a given point or area.

Path-finding The “Stroll” Javascript module was created to provide a high-level interface into computing a “nature score” for a given point or area.

Application

The final prototype featured the path finding functionality, as well as additional user-facing features such as the ability to record notes during your walk, capture places of interest, and customize your future routes based on past preferences and feedback.

Prototype Actual mobile application prototype, running during the demonstration at the Hackathon.

Winners The winning team photo!