I live in Los Angeles which just found out homelessness had increased 16% from last year, despite making cutting the numbers a priority. The homeless survey (PIT) is only performed yearly during 3 days in February. That's far too infrequent and it took 3 months to get relevant data.
What it does
Using image recognition, Surveyless allows a worker to take photos of areas with homeless tents and instantly get a count. This information in bundled into a Docusign report which they sign and forward to supervisors.
How I built it
I built and iOS app in Swift and used both the Docusign and Google Cloud APIs. Once I've captured an image, I run it through a Google visual analysis to identify tents. I cut these images out from the original and generate a document with additional information such as the location, count, etc. This is embedded into a Docusign envelope for signing and forwarding on through a workflow.
Challenges I ran into
Integrating the Google vision service required me to hand code the REST API as it's not part of their SDK. I also had to figure out the proper parameters to send, and had to parse out a large resulting JSON. The Docusign flow also was a little tricky as I'm trying to demo the flow with myself.
Accomplishments that I'm proud of
I got the whole process working and it's fairly easy to use.
What I learned
I learned a lot about the Google vision functions, and about composing Docusign documents.
What's next for Surveyless
The service needs engagement with the right civic entities to implement a real version.