Networking events can provide opportunities to expand your business contacts, increase your sphere of influence and attract new clients. The ability to remember names is an important skill for networking events. It can be difficult to build relationships with potential new customers if you can't remember their names and background. Remembrance allows you to avoid these embarrassing encounters by remembering for you.

What it does

Remembrance leverages computer vision algorithms to track and recognize previously met colleagues by taking a quick picture of them and entering their information. By including them in a frame of a shot, Remembrance will instantly recognize them and provide their details.

How we built it

Node/Express app built with a simple Bootstrap front-end and deployed on AWS. We used tracking.js to recognize faces through the camera and Clarifai's Predict and Train API to build and utilize a classification model.

Challenges we ran into

Clarifai's API limit (10 per second) was an issue since we were comparing every frame each time a face was tracked. We also faced some bugs with Clarifai's API which led to weird behavior in caching and retrieving existing concepts from a model.

Accomplishments that we're proud of

There is currently no support for live video tagging in Clarifai; our application is built it from the ground up by classifying video frames using their Image APIs. We got around multiple constraints such as request limits and certain API bugs by implementing our own rate limiting and caching algorithms.

What we learned

Coffee is good for the soul.

What's next for Remembrance

Although Remembrance can classify multiple faces in a frame at once, it currently can only support displaying a single person's information. Next step is to create a mechanism to intuitively display various contacts' details to the user.

We also plan to make Remembrance a standalone application. The vision is to make use of wearable devices, such as Snapchat Spectacles or Google Glass, to be able to recognize people through the user's eyes.

Built With

Share this project: