We were inspired by the story of the large and growing problem of stray, homeless, and missing pets, and the ways in which technology could be leveraged to solve it, by raising awareness, adding incentive, and exploiting data.
What it does
Pet Detective is first and foremost a chat bot, integrated into a Facebook page via messenger. The chatbot serves two user groups: pet owners that have recently lost their pets, and good Samaritans that would like to help by reporting. Moreover, Pet Detective provides monetary incentive for such people by collecting donations from happily served users. Pet detective provides the most convenient and hassle free user experience to both user bases. A simple virtual button generated by the chatbot allows the reporter to allow the bot to collect location data. In addition, the bot asks for a photo of the pet, and runs computer vision algorithms in order to determine several attributes and match factors. The bot then places a track on the dog, and continues to alert the owner about potential matches by sending images. In the case of a match, the service sets up a rendezvous with a trusted animal care partner. Finally, Pet Detective collects data on these transactions and reports and provides a data analytics platform to pet care partners.
How we built it
We used messenger developer integration to build the chatbot. We incorporated OpenCV to provide image segmentation in order to separate the dog from the background photo, and then used Google Cloud Vision service in order to extract features from the image. Our backends were built using Flask and Node.js, hosted on Google App Engine and Heroku, configured as microservices. For the data visualization, we used D3.js.
Challenges we ran into
Finding the write DB for our uses was challenging, as well as setting up and employing the cloud platform. Getting the chatbot to be reliable was also challenging.
Accomplishments that we're proud of
We are proud of a product that has real potential to do positive change, as well as the look and feel of the analytics platform (although we still need to add much more there). We are proud of balancing 4 services efficiently, and like our clever name/logo.
What we learned
We learned a few new technologies and algorithms, including image segmentation, and some Google cloud platform instances. We also learned that NoSQL databases are the way to go for hackathons and speed prototyping.
What's next for Pet Detective
We want to expand the capabilities of our analytics platform and partner with pet and animal businesses and providers in order to integrate the bot service into many different Facebook pages and websites.