INTRODUCTION

In densely populated areas such as urban environments, public events, and transportation hubs, monitoring crowd behavior is paramount for ensuring public safety and security. Traditional methods of surveillance and crowd management often rely on manual monitoring, which can be labor-intensive, time-consuming, and susceptible to human error. Moreover, as the complexity and scale of modern urban environments increase, the need for more sophisticated and efficient crowd monitoring solutions becomes increasingly pressing.

PROBLEM STATEMENT

The project aims to address the challenge of effectively monitoring crowded environments by developing a computer vision system capable of autonomously detecting anomalies in crowd behavior. This involves creating robust algorithms to analyze live or recorded video feeds, distinguishing normal crowd patterns from abnormal events such as fights, accidents, or suspicious behavior, and facilitating timely interventions to ensure public safety and security.

ABOUT THE DATASET

The mall dataset was collected from a publicly accessible webcam for crowd counting and profiling research. Video length: 2000 frames Frame size: 640x480 Frame rate: < 2 Hz The dataset is composed by RGB images of frames in a video (as inputs) and the object counting on every frame, this is the number of pedestrians (object) in the image. The images are 480x640 pixels at 3 channels of the same spot recorded by a webcam in a mall but it has different number of person on every frame, is a problem of crowd countinG.

RESULTS