People Counting System Using Python

Nowadays population growth increased. The linear growth of population has resulted in a large number of people comes to public places. This system will thus provide the count of the person in particular area malls, supermarket, etc. some business is only depending upon on the customer schedule and its timing. Hence, our work satisfies the problem and provides a solution by designing a people counting system. So, the single shot detector (SSD) mobile net is proposed along with a centroid tracker. This model replaces the VGG16 base network with a Mobile Net network for better extracting features, and after the base network, it connects with six convolutional layers for classification. The centroid tracking algorithm obtains bounding box coordinates from an object detector SSD and applies them to calculate the center of a bounding box. After the centroid is calculated, it will assign an id to each person and this model operates the dataset with training and testing the data which is shown in fig. 4. The maximum true positive rate (TPR) is 95.03% and FPR (false positive rate) is 0.08%. A total of 2416 positive images samples and 1218 negative samples. After that, the accuracy of the model is obtained and also 96.64% accuracy rate is obtained by testing the given data set. Sometimes method do not produce high accuracy due to the high traffic counting, clothes of person, shadow issue etc.