-
-
Mosey Dashboard
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Inspiration
Inspired by Jane Jacobs' 1960's revolutionary book "The Death and Life of Great American Cities" , the term "walkability" sprung to life. Generally, walkability is a measure of "how walking-friendly an area is". Walkability is positively associated with property values, social capital and fulfillment, public health outcomes, and environmental sustainability. Yet in the roughly 60 years in which people have been measuring walkability, the factors of interest have stayed roughly the same: coverage of sidewalks, pedestrian safety, proximity of housing to businesses and schools, etc.
These factors address the question of whether a person could realistically walk in a neighborhood, but to truly get more people out of cars and onto their feet, we must measure factors of walkability that address the question of whether a neighborhood is “walking-friendly.” In other words, will people have the amenities and resources they need to actually enjoy walking?
That's where Mosey comes in
What it does
Mosey is a machine learning-driven interactive Esri tool which improves the accessibility of pedestrian-friendly streetscapes for healthier, more sustainable neighborhoods. Mosey processes Google Street View (GSV) data to recognize where there are core amenities for enjoyable walkability and maps the locations of these amenities in an interactive dashboard. For the purpose of the hackathon, Mosey will demonstrate this capability for trash and recycling bins, which have one of the strongest positive relationships with pedestrian activity of all urban design features (citation). Mosey is designed for use by the public both to see where bins are currently, but also to crowdsource requests for new bins (or other amenities). These crowdsourced data points draw in a second audience for Mosey, public officials, who can use the dashboard to identify potential locations for new urban pedestrian amenities.
Street data is obtained from The United States' Census Bureau 2021 TIGER/Line® Shapefiles
How we built it
Mosey is built using ArcGIS Pro tools in tandem with standard Python data science libraries such as numpy, pandas, matplotlib and Tensorflow. The Google Street View API code accepts road coordinates to generate street images (5500+) for a city of interest (Harrisburg, PA for our analysis). These images are then fed into a Convolutional Neural Network specially designed in TensorFlow for our classification task of identifying public garbage/recycling bins. ArcGIS Experience Builder is used for interaction with the public in order to survey first-hand what individuals think of their surroundings with regards to garbage/recycling bins as well as cleanliness. For presenting our results to public officials and judges, we utilized ArcGIS Dashboards to create live maps and updates as to the status and locations of public garbage and recycling bins.
Challenges we ran into
Our team ran into many challenges and we were able to overcome most of them through teamwork, time management, and allowing individuals to contribute their greatest strengths. Some of the largest challenges that we faced was learning to create clean data which could be fed directly into the Google Street View API to generate training and testing images for our machine learning model. In addition, creating and training a Convolutional Neural Network (CNN) over a weekend was a rather daunting task. A large quantity of clean, trained data was necessary which took a lot of hours for our team to gather. Ultimately the CNN converged to a training accuracy of 77.03% which set a great baseline for our project and justified our hard work. We are certain that our model can be improved with the addition of more data and more time!
Accomplishments that we're proud of
The way that our team managed to deal with stress, time constraints, and coding challenges despite working entirely remotely across the globe is our greatest source of pride. We chose a very technical subject, creating a machine learning model over the span of a weekend was incredibly challenging but nonetheless rewarding. We understand that our model needs work, but we are very proud that we have a working model which can be further improved upon.
What we learned
We learned how to work in teams remotely and how to collaborate across many time zones. All of us learned a little bit about working with the Google Street View API as well as python programming. For the members of our team that were not familiar with Artificial Intelligence, a fair amount of Deep Learning was learned in order to get an understanding of how the algorithm we designed is working "under the hood".
What's next for Mosey
Mosey has so much work to do. We are just getting started. Mosey can easily improve by generating more training/testing data for its convolutional neural network technology. We would also like to see a generalization of our workflow for any city which is interested in improving its walkability and street-friendliness!
Log in or sign up for Devpost to join the conversation.