Sparse to Dense Scale Prediction for Crowd Couting in High Density Crowds
- PDF / 4,474,308 Bytes
- 15 Pages / 595.276 x 790.866 pts Page_size
- 23 Downloads / 298 Views
RESEARCH ARTICLE-COMPUTER ENGINEERING AND COMPUTER SCIENCE
Sparse to Dense Scale Prediction for Crowd Couting in High Density Crowds Sultan Daud Khan1
· Saleh Basalamah2
Received: 9 April 2020 / Accepted: 28 September 2020 © King Fahd University of Petroleum & Minerals 2020
Abstract Head detection-based crowd counting is of great importance and serves as a preprocessing step in many visual applications, for example, counting, tracking, and crowd dynamics understanding. Despite significant importance, limited amount of work is reported in the literature to detect human heads in high-density crowds. The problem of detecting heads in crowded scenes is challenging due to significant scale variations in the scene. In this paper, we tackle this problem by exploiting contextual constraints offer by the crowded scenes. For this purpose, we propose two networks, i.e., sparse-scale convolutional neural network (SS-CNN) and dense-scale convolutional neural network (DS-CNN). SS-CNN detects human heads with coarse information about the scales in the image. DS-CNN utilizes detection obtained from SS-CNN and generates dense scalemap by globally reasoning the coarse scales of detections obtained from SS-CNN via Markov Random Field (MRF). The dense scalemap has unique property that it captures all scale variations in image and provides an aid in generating scale-aware proposals. We evaluated our framework on three challenging state-of-the-art datasets, i.e., UCF-QNRF, WorldExpo’10, and UCF_CC_50. Experiment results show that proposed framework outperforms existing state-of-the-art methods. Keywords Crowd counting · Head detection · High-density crowds · Crowd analysis
1 Introduction Ensuring crowd safety and providing security to the participants of mass events is challenging problem and receiving great attention from the scientific community. With the growing population and increasing urbanization, mass events like marathons, sports, religious festivals, concerts, and carnivals organized frequently. In order to ensure crowd safety and security at these mass events, adequate safety measures must be adopted by the event organizers and security personnel. Crowd disasters still occur frequently, for example, during Love Parade [1] and Hajj [2], despite all safety measures. Crowd disasters usually attribute to critically high densities in a constrained environment. To avoid crowd disasters and in order to ensure crowd safety, it is important to analyze crowd dynamics. Understanding crowd dynamics has
B
Sultan Daud Khan [email protected] Saleh Basalamah [email protected]
1
National University of Technology, Islamabad, Pakistan
2
Umm Al-Qura University, Makkah, Saudi Arabia
numerous applications, for example, anomaly detection [3– 6], congestion detection [7], crowd counting, tracking [8,9] and many others. Among these applications, crowd counting has achieved tremendous attention from the computer vision community during recent years [10–14] The goal of crowd counting is to estimate total number of pedestrians in
Data Loading...
