Contactless Checkout: Reduce Checkout Time
Client
Multinational chain of convenience store based in Arizona
Business Problem
For a shopper, nothing can be more annoying than waiting in a long line for a single sales associate at a single open register
Objective
Speed up the retail checkout process by reducing the checkout time
Abzooba’s Solution:
- Built a deep learning solution to identify the products on the checkout counter
- A couple of cameras, adjacent to the counter, placed perpendicularly to each other to prevent occlusion, provides input to the system
- Object detection is done using TensorFlow for 10 classes
- A deep neural net comprised of Faster-RCNN with ResNet101 based model is being used to detect SKU instances
- The output is directly provided to the billing software to generate an automated invoice
- The sales associate can also validate the same before the invoice is generated
Business Benefits:
- Augments the checkout process, reducing the checkout time by almost 45%
- Detection accuracy of 93%