Hybrid Approach for Retail Store Auditing Using CRNN

In today's world, On-Shelf Availability (OSA) is critical in the field of selling and marketing. It is an important measure of performance for the retail sector that has a significant impact on profit and customer retention. Out-of-Stock causes sales to decline and customer retention to dwindle. Because of the firm's fierce competition, vendors and providers place attractive offers to please customers and keep them returning to their shops. In this work, a hybrid framework has been proposed to prevent out-of-stock problems by utilizing IoT and deep learning-based models for product label detection and recognition. To extract objects and their features a recurrent convolutional neural network has been utilized. IoT-based sensors help identify minimum loads on the shelf. The proposed approach is linked to a camera apparatus which detects inventory levels on the racks and generates signals to the store manager if items are removed from the rack or misaligned. The experimental results show that by using the proposed approach, it is feasible to recognize front-facing items and empty content on the rack with an accuracy ranging from 83%-99%.