A smart object detection for grocery stores and super markets which detectes if the products and under-stock, out of stock or misplaced.
Detection was done by fine tuning a pre-trained YOLOv8 model on a 110k SKU dataset, which gave it capabilities to achieve 98% accuracy on detecting any type of product
Classification was optimized by extracting the flattened feature vector using Inceptionv3 CNN model and comparing it with sorted feature vectors dictionary of the inventory.
the average of the maximum and minimum limit of the product will create a dictionary of the limits and the products will be added to this dictionary based on the center coordinates. which also saves the process of seperately trainig of rack identification
Kaushal G / @kaushalg47