PavedWay

• 

  app (mockup)

• 

  coverage map

• 

  tactile paving connectivity

Inspiration

What it does

Provide coverage of tactile paving in cities and guidance for visually impaired persons in unfamiliar places.

How we built it

Built a Deep Neural Network for to automatically detect tactile pavings from satellite images and connected the pathways using Graphs analysis.

Challenges we ran into

• No available labelled data
• Partial satellite coverage
• Poor quality data

Accomplishments that we're proud of

Proof of concept that is end to end: from scratch to a new competitive guidance tool.

What we learned

The complexity of traveling into new places for blind peoples, and a new way of solving problems from scratch.

What's next for PavedWay

Run it on more data (other cities) connect our network with a routing services

Built With

• pygsp
• python
• pytorch