Tell us about your project!
Retail
Product classification and images annotation helped to improve grocery store efficiency by 40%
Training data for AI and machine learning
About case:
About case:
Industry and use case
Retail, store shelves monitoring
DATA
100,000 shelves images with products
Project duration
4 months
Industry and use case
Retail, store shelves monitoring
DATA
100,000 shelves images with products
Project duration
4 months
Challenge:
Challenge:
The customer wanted to automate the process of grocery store shelf monitoring, including automatic product identification using a neural network. The goal was to improve the merchandising strategies and analyze the impact of promotion campaigns on sales in real-time. The main challenge for data labelling and classification resulted from a wide range of product categories, types, and packages.
Solution:
Solution:
The Training Data annotators were split into two teams. One group prepared product requirements by looking for product examples in each product category. The second group used the requirements to label actual shelves images. This approach allowed the team to reach high accuracy of product classification.
Outcomes:
Outcomes:
in cost savings
Accurate real-time data about product merchandising and sales
Up to 40%
Feedback
Training Data organized the process most efficiently when two teams worked on data preparation and labelling in parallel. We will definitely continue working with the team.
Roman O. Project Manager