2023-10-05
AI and machine learning to meticulously analyze the shapes of products on a shelf and ensure precise counting
- Customer
- Boutique Rugs LLC
- Project manager on the customer side
-
- Year of project completion
-
2023
- Project timeline
-
July, 2023 - September, 2023
- Project scope
-
250 man-hours
- Goals
-
The process of taking inventory in our warehouse was consuming excessive time and manpower, leading to inaccurate results due to the susceptibility of manual counting errors when dealing with dozens of products on a shelf. In response, we created a mobile application that swiftly counts our products within seconds by capturing a snapshot of the shelf.
Rug Counter stands as an in-house software application that employs AI and machine learning to meticulously analyze the shapes of products on a shelf and ensure precise counting. We undertook not only the development of the code but also the training of the algorithm to yield accurate results.
- Project Results
-
Rug Counter has proven instrumental in several aspects, including:
- Precision: Our inventory information is now more accurate.
- Cost-effectiveness: The need for a reduced staff to count inventory has enhanced efficiency.
- Swiftness: Previously taking weeks, the inventory counting process in our warehouse has been significantly shortened to a matter of days.
The uniqueness of the project
While there were other counting applications available, they faced a limitation – they could only accurately count precisely round shapes such as tree stumps or pipes. As our products are stacked in diverse arrangements, their shapes may vary. Rug Counter overcomes this challenge by analyzing these altered product shapes, ensuring precise results. Moreover, it seamlessly integrates with our warehouse management system, updating inventories on both the WMS and various e-commerce platforms and marketplaces.
- Used software
-
Software
Java, Python, OpenCV, Android Studio, Php WebServer
Equipment
Android Device
Auxiliary Systems
OpenCV image manipulation technology
AI Algorithms
- Difficulty of implementation
-
The shapes of the products in the stack may vary. To ensure precise outcomes, we had to train the AI using more than 100 images.
- Project Description
-
In the past, our warehouse team engaged in manual inventory counting, a time-consuming process fraught with uncertainty, especially when dealing with over a hundred rugs that were not uniformly aligned. To streamline this operation, we developed a mobile application compatible with both mobile scanners and tablets, enabling warehouse staff to swiftly count products on stacks within seconds.
This involves the staff capturing images of the piles, masking undesirable portions in black, and utilizing AI for analysis. The AI examines the images, identifying and notifying the user of the recognized rug areas. In cases where the AI overlooks certain rugs, users can manually mark those items, contributing to the training of the AI for improved accuracy.
Following the completion of the counting process, Rug Counter seamlessly integrates with our warehouse management system. This integration not only updates the inventory within the warehouse management system but also synchronizes across all our sales channels, including marketplaces and ecommerce platforms.
The video at the link demonstrates how the application works: https://youtu.be/V4QyL9GA47Q
- Project geography
-
We are using the app in our warehouse (Türkiye)
- Additional presentations:
-
Rug Counter.pdf
