Inspiration

The deportation and forced separation of immigrant families crossing into the United States has psychological effects on individuals and families and gives rise to a public health crisis that can affect entire communities, according to a new report published in the American Journal of Community Psychology. Children who lose a parent to sudden, forced deportation experience anxiety, anger, aggression, withdrawal, a heightened sense of fear, eating and sleeping disturbances, isolation, trauma, and depression. Children also experience housing instability, academic withdrawal, and family dissolution. Older children often need to take on jobs to help support the family. With families being separated by raids and children being inhumanely kept in detention centers or sent to far-flung locations, our team found it necessary to address this social injustice.

What it does

Reunited provides a platform for families to find their missing children as well as a way for concerned citizens to report a child's information to help find their family. The web app compares two photographs of children and determines whether those two children are the same person with an accuracy of over 95%. This creates a match between parent and temporary caretakers in order to reconnect the children with their families.

How we built it

We used Amazon Web Service's Rekognition API to determine identity of the child regardless of factors such as lighting, time, angle, environment, or superficial changes like hair color. We also created a database using mySQL and hosted on AWS where we save all the images and the information about the missing children and their parents. We put our algorithm and everything together via a flask web app where we built the backend using REST APIS in python and used SQL to deal with the database.

Challenges we ran into

We tried to code our own feature and use Naive bases algorithm to identify the children in addition to the existing free api's. We made good progress and reached an okay accuracy, however, we ended up using the amazon Rekognition api as its accuracy was very high compared to ours. The initial API's we encountered for facial recognition or image classification were insufficient and inaccurate in identifying age and gender. It almost made us change our idea for the hackathon when fortunately we found AWS's Rekognition API. Furthermore, we had difficulties connecting our back-end to our front-end. We also had trouble finding a way to store images into our database.

Accomplishments that we're proud of

We are very proud to have created an MVP within a little over a day, and having overcome obstacles and difficulties along the way. We are very proud that we were ably to successfully create the database with the images stored with only one team member having experience in databases and AWS. We are also proud with the progress we made with coding our feature in which we ended up not using. One of our team members too an AI class and she was thrilled that could actually code such a complicated feature in short time and use her class experience in the hackathon even if we didn't use in the project. We are also proud that we took into consideration the security factor and accounted for not presenting all the children information as this might have negative consequences on the children and the parent.

What we learned

The front-end members of the team learned some React.js in an attempt to incorporate some components into the web app. We also learned a lot about working with Amazon Web Services, which only one in the team was proficient in prior to this hack.

What's next for Reunited

We hope to extend our scope to include any immigration or refugee crisis in the United States, not just children. Furthermore, the ability to take a picture within the app instead of simply uploading one would significantly improve user experience. We hadn't have enough time to include the live image analysis instead of uploading picture but we seriously think of implementing it after the hackathon. Finally, we plan on deploying the app so everyone has access to them.

Share this project:
×

Updates