Abstract

A Population growth is rising in modern times. Due to the population's linear increase, many individuals now frequent public spaces. Thus, this technique will give the number of people in a specific area's malls, supermarkets, etc. Some businesses exclusively depend on the timing and schedule of their customers. Therefore, by creating a system for counting people, our work satisfies the problem and offers a solution. Therefore, a mobile single shot detector (SSD) network and centroid tracker are presented. For better feature extraction, this model swaps out the VGG16 base network for a Mobile Network, and for classification, it links with six convolutional layers following the base network. In order to calculate the centre of a bounding box, the centroid tracking algorithm uses bounding box coordinates from an object detector SSD. Each person will receive an ID when the centroid is calculated, and this operates the dataset with training and testing the data.