ADAP

Ever wondered how much effort goes into creating new paths for both walking and cycling? Take a picture of a path and find out more!

Comment

Inspiration

Pedestrians, cyclists and PMDs. With paths in Singapore getting ever-more crowded along with Singapore's push toward last-mile transit, the methodologies by which Singapore creates these public infrastructures are currently still excruciatingly manual.

As such, our main inspiration for this project was the automation of this manual process by leveraging the latest advancements in capabilities and accessibility within the field of Artificial Intelligence to harness publically available Big Data to create a solution that effectively contributes to the Urban Living domain of Singapore's Smart Nation initiative. To solve this issue, we are proud to present our tech-enabled solution: ADAP.

What it does

ADAP encapsulates 3 separate phases: Assessment, Data visualization, and Urban Planning.

The first phase, Assessment, analyzes pictures taken of paths in Singapore and location coordinates to generate a walkability and bikeability score for that particular path in Singapore.

The second phase, Data Visualization, Mapviews is utilized to display the walkability and bikeability scores of regions of interest, both for the existing state and future plans. To abstract away unnecessary details, the accumulated path locations have their scorings represented by different color icons on spread out on an interactive Singapore map.

The last phase, Urban Planning, enables government agencies to gain an overview of the walkability and bikeability scores of current infrastructures around Singapore. This highlights potential regions that may need additional work for authorities to take note. Should improvement works be carried out, new images can be provided to update the scores of involved regions. Additionally, the mobile app allows design contractors to generate on-demand walk and bikeability surveys simply from their initial designs. This improves the effectiveness of the planning phase without requiring any ground works.

How we built it

Our AI model consists of two key components: an Artificial Neural Network (ANN) and an image segmentation model. First, our image segmentation model (Google DeepLab v3+) is trained upon the Cityscapes Dataset, as well as tested on San Francisco and Singapore urban landscapes to provide us with segmented images. This data is then fed into our ANN, along with other meaningful web-scrapped data that we collected, to generate a walk score and bike score based on the input of a particular path in Singapore.

Our front-end app, meant for urban planners and on-site surveys, is built using React Native to provide an accessible means to query our AI model to extract on-demand scores per user requirements. This app will also provide urban planners the capability to upload artist impressions and mockups to evaluate their walkability and bikeability scores.

Challenges we ran into

  • Difficulty in finding meaningful data sources to train our ANN model on
  • Difficulty in the integration of multi-faceted components of the project

Accomplishments that we're proud of

  • AI model is able to generate meaningful and accurate scores based on the information given
  • Querying of data through API calls works smoothly to give intended outputs with relatively low latency
  • Front-end works well to provide a smooth user experience during usage

What we learned

Issue resolution was a key aspect that we picked up. Over the last 60 hours, many, many issues popped up and our team was fortunately able to resolve these issues through effective communication.

What's next for ADAP

We hope to create a more robust application with lower latency in order to truly satisfy on-demand usage. We also hope to draw upon more datasets in order to provide an even-more minute and accurate solution that will suit the needs of urban planners.

Appendix

-Rattan, A. (2012). Modelling Walkability. Modeling Walkability. Retrieved June 12, 2022, from https://www.esri.com/news/arcuser/0112/modeling-walkability.html

-Ito, K., & Biljecki, F. (2021). Assessing bikeability with street view imagery and Computer Vision. Transportation Research Part C: Emerging Technologies, 132, 103371. https://doi.org/10.1016/j.trc.2021.103371

-Nagata, S., Nakaya, T., Hanibuchi, T., Amagasa, S., Kikuchi, H., & Inoue, S. (2020). Objective scoring of streetscape walkability related to leisure walking: Statistical Modeling Approach with Semantic Segmentation of Google Street View images. Health & Place, 66, 102428. https://doi.org/10.1016/j.healthplace.2020.102428

-Markevych, I., Schoierer, J., Hartig, T., Chudnovsky, A., Hystad, P., Dzhambov, A. M., de Vries, S., Triguero-Mas, M., Brauer, M., Nieuwenhuijsen, M. J., Lupp, G., Richardson, E. A., Astell-Burt, T., Dimitrova, D., Feng, X., Sadeh, M., Standl, M., Heinrich, J., & Fuertes, E. (2017). Exploring pathways linking greenspace to health: Theoretical and methodological guidance. Environmental Research, 158, 301–317. https://doi.org/10.1016/j.envres.2017.06.028

Built With

Share this project:

Updates