Tap Placement | Devpost

Output image (small)
Edge detection
Fully mapped village
Input image

Challenges applied to

BMI challenge
Investec challenge
BCS challenge

Inspiration

We have all worked with the innovative charity eWaterPay. They work installing sustainable water systems in rural African villages. Recently they have encountered a challenge in surveying potential sites for water systems. They have to spend weeks surveying villages to find population density and attempt to find the best position to install water taps, given limited resources.

What it does

Our system uses computer vision to locate rooftops in rural areas. It then calculates the size of each home and uses this to estimate a population. The position and population of each house is then fed into our decision algorithm which calculates the optimum position for a given number of taps.

How we built it

We developed a frontend webapp using Angular to allow the user to select a location using Google maps. We used the flask framework to develop an API to allow the client to send the location which is then analysed by the backend. The backend uses complex computer-vision and decision mathamatics to calculate optimum placements and return this to a user.

Challenges we ran into

It was initially difficult to determine the exact position of rooftops, but with rapid alterations to our algorithms we developed a more accurate estimate. Again, our inital tap placement algorithm was exceptionally slow, but we brainstormed overnight to reduce this from polynomial time to linear time.

Accomplishments that we're proud of

We have never worked on computer vision before and we are so proud of how accurate it has turned out to be. And the results we are getting from tap placements seem almost human.