ARC: A Vision-based Automatic Retail Checkout System
Authors:
Syed Talha Bukhari,
Abdul Wahab Amin,
Muhammad Abdullah Naveed,
Muhammad Rzi Abbas
Abstract:
Retail checkout systems employed at supermarkets primarily rely on barcode scanners, with some utilizing QR codes, to identify the items being purchased. These methods are time-consuming in practice, require a certain level of human supervision, and involve waiting in long queues. In this regard, we propose a system, that we call ARC, which aims at making the process of check-out at retail store c…
▽ More
Retail checkout systems employed at supermarkets primarily rely on barcode scanners, with some utilizing QR codes, to identify the items being purchased. These methods are time-consuming in practice, require a certain level of human supervision, and involve waiting in long queues. In this regard, we propose a system, that we call ARC, which aims at making the process of check-out at retail store counters faster, autonomous, and more convenient, while reducing dependency on a human operator. The approach makes use of a computer vision-based system, with a Convolutional Neural Network at its core, which scans objects placed beneath a webcam for identification. To evaluate the proposed system, we curated an image dataset of one-hundred local retail items of various categories. Within the given assumptions and considerations, the system achieves a reasonable test-time accuracy, pointing towards an ambitious future for the proposed setup. The project code and the dataset are made publicly available.
△ Less
Submitted 17 May, 2021; v1 submitted 6 April, 2021;
originally announced April 2021.
