eCashier

Welcome Page
Face Detection
Face Recognition
Item QR Code Scan
Live Monitoring

Inspiration

The payment system right now is very inefficient ,slow and tedious. Hence, we hope we can improve this system using artificial intelligence.

What it does

It can recognize customer's identification using face detection and recognition algorithm, and identify item's information by recognizing QR code through a single camera. Finally, a transaction record will be generated and processed unlike traditional banking system, but through a futuristic payment method.

How we built it

For the machine learning part, we used Resnet neural network which is formed by 29 layers. For the back-end development, we used MySQL as the database, Java Spark as MVC, and Jade as template. For the front-end, we used HTML5 to achieve livestreaming, track.js to detect customer's face, and instascan and Google's ZXing to recognize item's QR code.

Challenges we ran into

We tried different ways to recognize item's information, and finally we chose QR code because of the limitation of hardware. Integrate different frameworks, languages, and technologies.

Accomplishments that we're proud of

Finally it works very well. The detection and recognition accuracy can be surprisingly high, like 98%.

What we learned

Image processing, computer vision, front-end, back-end development, and Agile Developing Model.

What's next for eCashier

Instead of scanning QR code, we may improve object recognition method with high accuracy rate and fast speed. Provide more convenient functions and benefits to customers, such as promotion. track.js using ViolaJones Algorithm which is not very efficient for face detection, we will replace it with opencv in future.