Public space is a place that generally used by the community in order to meet their needs and where crowds are usually formed. In a crowd, the number of people can be the first indicator in an anomaly in where the more number of people exists in a crowd, the more supervision is needed on the crowd to prevent chaos or other things that are not desirable in public spaces. The need for a crowd-counting is certainly needed to facilitate supervision and also the awareness of people in crowds. This research meant to develop a system that can identifies a crowd based on the number of people that exist in the crowd and also give the number of people as an output. The system applied the Convolutional Neural Network (CNN) algorithm. The CNN model is trained using a labeled crowd dataset with a total of 4372 crowd photos. The CNN works as a regression model that will count the number of people from the feature extracted from the image. The evalution shows the Mean Absolute Error value achieved is 55.1176 in the test data.
Copyrights © 2023