Overview / Usage

Our Product has three modules to it.

To detect and identify selected products realtime accurately.

To detect the empty slots in the racks and to identify the misfits. i.e. Pampers cannot go into the rack where Tide is arranged. We could accurately identify the objects, find the empty slots if any in the rack and we also spot the wrongly fit items into racks.This help us in smart space optimization, accommodate more items in the shelves appropriately.

This also helps the store keepers in refilling the items in time to avoid customers facing challenges. We also can detect the empty racks which woul help the store keeper in refilling it!

To detect the dimension of the chosen products and to make the billing fully automated - No barcode required at any point for the billing now. Implementation - The implementation is out and out with the Deep learning concepts and the Fast RCNN was found the best fit. We get close to 100% accuracy for both the modules.

​

We request you to have a look @ the Demo Video. https://youtu.be/UkKANGUcCvE

Methodology / Approach

Our system’s backend is deployed on Azure Cloud. The backend takes the input stream from the site camera deployed at the billing station. All the processing including object detection, localization, and dimension estimation is completely done in Linux VM on Azure Cloud. Used Compute optimized Fsv2 instance with 2vCPU and 4GB RAM. We also used Image Detection - Computer Vision API of Azure cognitive services for enhancing object detection. The billing data after the entire billing process is completed, is stored in MariaDB which is a relational database and a lot of analytics can be given from this stored data.

Technologies Used

Intel Python

AI

​

Collaborators