Convolutional-Neural Network-Based Image Crowd Counting: Review, Categorization, Analysis, and Performance Evaluation

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Traditional crowd-counting methods are being replaced by AI and machine learning.
• This shift enables adaptive monitoring and control of dynamic crowd gatherings.
• Current challenges include occlusion, clutter, and scale variations in crowd images.

Purpose of the Study:

• To review, categorize, and analyze Convolutional Neural Network (CNN)-based crowd-counting techniques.
• To evaluate the performance of the latest CNN crowd-counting methods.
• To highlight potential applications and future research directions for CNN-based crowd counting.

Main Methods:

• Review and categorization of existing CNN-based crowd-counting literature.
• Detailed performance evaluation of selected CNN techniques.
• Analysis of limitations and distinctive features of various approaches.

Main Results:

• Convolutional Neural Networks (CNNs) are identified as a promising technology for intelligent crowd counting and analysis.
• A comprehensive evaluation of the latest CNN-based crowd-counting techniques is provided.
• Potential applications and future research avenues are discussed.

Conclusions:

• CNNs offer advanced capabilities for adaptive monitoring and management of crowd gatherings.
• Addressing challenges like occlusion and scale variation is crucial for improving CNN-based crowd counting.
• Future research should focus on enhancing CNN designs for robust crowd analysis, including smart city applications with the Internet of Things (IoT).