Artificial Intelligence in Product Counting

AI Case Study: Product Counting

Advances in computer vision make it possible to automate many tasks that require human presence for an assessment of situation and decision making. For example:

  • stock control and inventory management
  • detection of defects during manufacturing
  • or, more generally, analysing location and measuring size of objects on image or video frame

The purpose of this project was to develop the AI model that is capable of recognising, identifying and counting products on a shelf.

The below video demonstrates how model responds to product movement events in real time. When current level for each product is reaching a specific limit, critical level, an event is generated that automatically sends product replenishment requests.

If you have retail store located in Melbourne, Australia we would like to offer you a free pilot in using Artificial Intelligence to conduct automatic real-time monitoring of your shelf stock levels. Please do not hesitate to Contact us to discuss.

Skip to relevant parts of the video by selecting below.

0:00 Start of the recording, all products are accounted for, four of each type.
0:01 Rice Cracker #1 is removed, product count is changed from 3 to 4.
0:02 Rice Cracker #2 is removed, product count is changed from 3 to 2, reaching critical level. Product replenishment request is generated.
0:03 Rice Cracker #3 is removed, product count is changed from 2 to 1.
0:04 Rice Cracker #4 is removed, product count is changed from 1 to 0.
0:06 Caramello Koala #1 and #2 are removed, product count is changed from 4 to 2.
0:09 Hershey's #1 and #2 are removed, product count is changed from 4 to 2, reaching critical level. Product replenishment request is generated.
0:10 Caramello Koala #3 is removed, product count is changed from 2 to 1, reaching critical level. Product replenishment request is generated.
0:12 Caramello Koala #4 is removed, product count is changed from 1 to 0.
0:14 Hershey's #3 is removed, product count is changed from 2 to 1.
0:16 Hershey's #4 is removed, product count is changed from 1 to 0.
0:18 Soup #1 is removed, product count is changed from 4 to 3.
0:19 Soup #2 is removed, product count is changed from 3 to 2, reaching critical level. Product replenishment request is generated.
0:23 Soup #2 is added, product count is changed from 2 to 3, cancelling critical level.
0:24 Soup #1 is added, product count is changed from 3 to 4.
0:26 Caramello Koala #1 is added, product count is changed from 0 to 1.
0:28 Caramello Koala #2 is added, product count is changed from 1 to 2, cancelling critical level.
0:31 Caramello Koala #3 is added, product count is changed from 2 to 3.
0:33 Caramello Koala #4 is added, product count is changed from 3 to 4.
0:35 End of recording. Three products (Soup, Caramello Koala and Freddo) are fully stocked, two products (Hershey's and Rice Crackers) have zero count.


This AI case study has been conducted in Australia, Melbourne by FifthOcean and Nola, a foot traffic counter & visitor analytics solution for retailers, venues & events.