Loginly

We ran through multiple tests to get the models running at max efficiency which included constantly checking the logs
This is the live data of the metrics of one of the models running on Loginly
This is our storage account on GCP which goes along with the folder structure
Folder Structure
Architecture
Sign up page where the user inputs their name, email, and 10 photos
Login Page where user submits their photo
Example of the number of people in the dataset (Not including highlighted -> Tests)
An example of the login storage account of the different logins containing the base64 image string

Inspiration

Inspired by the inefficiency and inconvenience of passwords, Loginly transforms the traditional approach to login systems into a secure yet easily accessible login confirmation system.

What it does

Loginly is a secure login system that uses facial recognition software rather than a password to access a user’s account. Passwords, while widely used and normalized, are

Inefficient: in order to keep lots of accounts secure, you need to have different passwords for each account leading to either lots of painstaking memorization or needing another medium to hold all of your password information.
Insecure: many people who run into the problem to needing too many passwords end up reusing passwords, or if they do end up making lots of passwords they all get stored into a single space. Both options are susceptible to data breaches and hacking.

However, using Loginly, a photo of the user’s face provides a secure and encrypted login confirmation system that is easily accessed and cannot be lost.

How we built it

Loginly was created in several steps: front end development where users sign up or login, and the backend development of the facial recognition software as well as the system’s connection to a database to retrieve existing login information.

Backend

When designing the back end, we wanted to have the utmost security while also having the fastest speeds. Google cloud functions allowed for us to use intelligent AI libraries authored by Python while also integrating in a nice seamless front-end fashion. they also allow for intelligent integration with the database. The data is stored in an encrypted format open the GCP storage which prevents hackers to access private images. In order to access the database, one needs a specific key which further prevent security.

Challenges we ran into

Problems

When building the AI model for the facial recognition I first used face_recognition library which allowed me to quickly identify nature points on a face to detect the identity. the problem was when integrating this with the GCP function it resulted in an error because the library is not supported which is why I used LBHR from cv2 which is part of their face cascade classifier program. this allowed for a intelligent and accurate facial recognition. As of prelimerary tests the accuracy is upward of 98%.

Experience and Areas of Expertise

The biggest challenge we had with Loginly was that it was a project outside of our areas of expertise. Some of us had never done front end work before, while for others of us we had little experience with accessing and using databases. Our team also consisted of both experienced coders and those who've just begun their coding journey, so time had to be invested into not just doing the project but learning various concepts from the bottom up.

Communication and Team Building

Each one of our team members worked from separate parts of the globe, even having a 9 hour time zone difference between some of our team members.

Accomplishments that we're proud of

Each individual portion of the project took time to complete, but connecting everything into a completed system was the most satisfying accomplishment of our team’s project. Connecting user data from the web page to servers where the facial recognition software did its work to referencing the login page and successfully completing a login was amazing. Additionally, the team elements of working around time zone and experience barriers, and being able to organize our project fluidly was another huge success.

What we learned

Each one of us came into this hackathon with different skills and backgrounds; similarly, each of us will leave with our own unique experiences and takeaways. In the end, the challenges we faced amounted to an experience where each of us learned new coding skills and how to work together to overcome various barriers. Some of the skills we learned included:

front end development using frameworks like Bootstrap
web development and web requests
database management.

We also learned how to organize ourselves to overcome the time zone differences and designate tasks as well as how to communicate our ideas properly to get as much done as fast as possible.