Vision-Based Empty Shelf Detection in Retail with Real-Time Telegram Notifications for Efficient Restocking

Empty spaces and fewer items on shelves of stores and big marts often dissatisfy the customer by making the items unavailable when needed. Empty spaces and fewer items on shelves of stores and big marts often disappoint customers by making items unavailable when needed. This also reflects the commitment of store staff to their work. As a result, there is a decrease in sales and a breakdown of trust between sellers and customers. Object detection is used to identify empty spaces and shelves with fewer items. Commonly used algorithms for object detection include CNN, YOLO, and SSD. Large, freely available standard datasets such as Pascal (plate number 1) and Pascal (plate number 2) are utilized, each containing around 20 classes for shelf item detection. Items are labelled as 'Out of Stock' along with their names. This labelling helps visually represent the items. Object detection often requires GPUs and a webcam. The system has developed a dataset containing four classes of grocery items. The labels for the items have been derived from their respective images, with annotations stored in separate image files. The system has been trained using the YOLOv5 algorithm. The output, consisting of images showing empty shelves or low item counts, has been connected to the Telegram API to notify store staff to restock as needed, streamlining the restocking process. This versatile application can be used for inventory management, research, and development and can also be integrated with commercial retail stores, utilizing CCTV cameras for monitoring.