Take a picture of a card
Have a contact in seconds
+ We wanted to create a light-weight service that eliminates a tedious process. + We intentionally avoided complete automation, to allow for human connection. + We found that this project would be easily expandable to simple recognition (as far as the user is concerned) of other physical objects, as we intended to open source it to allow for the contribution of others.
What it does
+ Automatically creates a phone contact by sending an image over SMS. + Utilizes image recognition to extract a name and phone number from a physical object. + Creates a virtual contact card using processed data. + Uses Flask back-end to serve .vcf files (phone contact file) to the Twilio API to be sent back to the user.
How we built it
+ Wrote a python script to processes images and return text information, and used it to retrieve a name and phone number. + Wrote functionality for Twilio to send a message containing a contact card given a .vcf (vcard) file + Linked the two parts together, and added further functionality to send the user's information to the owner of the business card.
Challenges we ran into
+ Implementing natural language processing + Issues with sending local objects, as contact cards, over MMS through Twilio + Compatibility issues with Android
Accomplishments that we're proud of
+ Managed to keep service lightweight, with an average response time of around 10 seconds + Used python across the board. This repo is 100% python.
What we learned
+ It is much harder to implement natural language processing than expected
What's next for EasyLink
+ Better image to text conversion + Implementing natural language processing, in order to better determine what parts of a business card is the name, company, email, etc.