Automated Logo-based Product Separation using Machine Learning

Abstract

Logo detection of different brands play an important role in advertisement industry and also in sorting of different products .In this paper a logo based product separation is done using deep learning (R-CNN) model. We collect dataset of logos of different brands and trained them using transfer learning Approach. This transfer learning approach reduces number of images required for training and can be train on CPU with less memory (RAM) is used. Our contribution is to apply Reduce convolution network (RCNN) model on our custom data set of logos and get accurate result .In this study the CIFR10 image data set is used which contain 50,000 training images that used to train RCNN. After logo recognition we separate the product through conveyer belt with the help of separators. The servo motors are used with model MG-945 having torque of 180 degree in this project .All processing is performed on MATLAB 2019a and training is done on single-CPU with RAM 8GB and core i-5,3rd generation.