2024-11-04
GondolIA - Recognition of product withdrawal
- Customer
- Atacadão Dia a Dia
- Project manager on the customer side
-
- Year of project completion
-
2024
- Project timeline
-
January, 2024 - May, 2024
- Project scope
-
720 man-hours
- Goals
-
The project aims to implement a computer vision model to identify the removal of products from shelves in real time, with the aim of reducing short-term stock-outs, improving the visual presentation of aisles and reducing operating costs. The model uses Vertex AI, which classifies images captured by cameras in shop aisles, enabling more efficient product management and timely replenishment.
- Project Results
-
The project reduced short-term stock-outs and improved the visual management of products in the aisles. The implementation of the computer vision model increased operational efficiency and the speed of shelf replenishment, reducing operating costs. Real-time data also helped with strategic decision-making to maintain product availability.
The uniqueness of the project
The project's unique advantage lies in its application of advanced computer vision technology for real-time monitoring of product removal from retail shelves. Using cameras to analyze frames, the system transmits data to a dashboard, enabling immediate decision-making. This innovative approach accelerates product replenishment and prevents stock-outs, setting it apart from traditional solutions with its precision and seamless integration with Vertex AI and Google Sheets.
- Used software
-
The following systems and equipment were used: Google's Vertex AI for the computer vision model, cameras for capturing images in the aisles, Google Sheets for storing withdrawal data, Looker for visualisation on dashboards and AppSheet for monitoring and validating refuellings. The cameras and data flow via API were also key components for the project to work.
- Difficulty of implementation
-
Implementing the project involved technical challenges, such as setting up the cameras to capture the correct angles in the corridors and training the model to correctly classify different types of movement. Integrating systems such as Google Sheets, Looker and AppSheet also required adjustments to ensure the accuracy and real-time updating of the data. Calibrating the model to cope with variations in lighting and other environmental factors was also a critical issue.
- Project Description
-
The GondolIA project aims to monitor and identify product removals from shelves using a computer vision model from Vertex AI. The cameras capture the aisle flow and send the frames to the API, which processes and classifies the actions as removal, non-removal or replacement. The result of the analysis is recorded in Google Sheets and visualised in Looker, providing real-time data for the replenishment team, which uses AppSheet to validate the need for replenishment. The process includes continuous monitoring of the aisles, with data categorised by sector, branch and more, for efficient management and reduction of stock-outs.
- Project geography
-
The project was implemented in physical store aisles, covering the monitored sectors and collecting specific data by branch and sector. This approach allowed for broad geographic coverage within the units, ensuring adequate reach for efficient inventory management and replenishment in strategic locations within the store.
