SUpost Watson

A web app that uses IBM Watson visual recognition to automatically classify photos

Comment

Inspiration

We envision our app as a "Third Eye" smart assistant -- live for use at Gerbil.tech. Through cognitive computing, enabled through IBM Watson Visual Recognition, Gerbil helps you sense the world around you, gain understanding and take actions in the world to help you.

We built a proof of concept app at Gerbil.tech to allow people to use our app to take a photo, recognize objects and quickly classify and add tags to describe those items. Integrated into our SUpost iPhone app (an online marketplace for Stanford students that gets 500,000 page views per month), the "Third Eye" smart assistant technology makes it easier and faster to advertise and sell your stuff.

User story

A Stanford student, Aaron, is graduating in June and wants to sell all the stuff in his room. Because Aaron will be moving back to Boston, he has a lot of stuff he needs to sell. Aaron opens the Gerbil app and snaps pictures of his room, including a photo his bike. Gerbil app automatically recognizes the photo as a bike. Gerbil publishes the photo into the correct category on the popular classifieds marketplace at Stanford, SUpost.com. Gerbil adds searchable tags to the bike photo to make the item more easily discovered by buyers and increases the likelihood that it will be found and sold. Aaron spends only 5 minutes using Gerbil to take photos of stuff in his room, and he's getting emails from buyers within 30 minutes. All his stuff is sold within a day, and he's ready to head home to Boston with an extra $2,000 in his pocket!

What it does

Uses Watson Visual Recognition to automatically classify and label images as either cars, bikes or neither. It also automatically generates tags that are related to the image.

How we built it

We compiled three folders of images from SUpost.com to train our classifier. One folder contained bikes, one contained cars and one contained a variety of images that were neither bikes nor cars. We then made a flask application that calls the Watson Visual Recognition API.

Next Steps

For this hackathon, we decided to classify just bikes and cars, but next steps would definitely include classifying different categories, e.g., laptops. We also think automatically tagging images would be useful for social media sites and other areas users store lots of images, making it much easier to organize and search for images. We would also like to expand to automatically tag not only images but video and audio as well. We envision that in 10 years, Gerbil.tech will evolve to become your "Third Eye" smart assistant to help you understand the world around you and take actions to improve your life.

Share this project:

Updates