Automatic Checkout System

Inspiration

It's no surprise the world is becoming more automated, but when I go to checkout my items in a store, I find it inconvenient to search for the barcode on every item that I have. So, this computer vision based checkout system solves that problem.

What it does

Detects any items in frame and adds it to the users cart which they can then checkout (checkout functionality not yet introduced). Aims to solve the problem of inefficient and frustrating barcode checkouts.

How we built it

Using tensorflow js pretrained cocossd model, and html made this simple website that serves as an automatic checkout system.

Challenges we ran into

Integrating the model into javascript, being able to use the webcam to stream images to the pretrained model.

Accomplishments that we're proud of

The final output works, which is something I am very proud of.

What we learned

I learned that the potential for computer vision based applications in industry is limitless.

What's next for Automatic Checkout System

Rather than using a pretrained model, collect a custom image dataset for a store if I had more time and resources to make a store specific checkout system.

Also use actual prices in a stores database, as I have currently set every item as $5.99, which is unrealistic and not practical, but with more time this would be simple to incorporate.