Identification of Number of Shop Visitors Using Faster Regions Convolutional Neural Networks

Abstract

The number of store visitors is essential data for the evaluation of a store; data collection become complicated if there are a large number of visitors. Object detection is one part of computer vision and digital image processing, which has the purpose of detecting a semantic object from a particular class. One approach is the Faster Region Convolutional Neural Networks (Faster R-CNN) method, which can be used to identify the number of shop visitors. This study used Faster R-CNN to calculate the number of store visitors. The process starts by retraining the Faster-RCNN model that has been trained before using the COCO dataset. Then it used the INRIA Person Dataset that has been labeled manually using Labelimg. After training and configuration modifications, the accuracy reaches 69% with real-time accuracy 63%.