Abstract
In machine vision, the tasks of both the crowd counting and the crowd
density estimation attracted a number of researchers from different
research institutes throughout the world. These tasks are of
significance importance while analyzing the surveillance videos and
images both at the real and offline modes. A number of different
conventional and modern algorithms are being applied to estimate the
accurate amount of people from the crowd from the target image or video
file but failed to provide the required results due to heterogeneity of
the data instances. To resolve the issue, machine learning models like
deep networks like Convolution Neural Network (CNN) and its variants
were introduced. In this paper, a detailed and comprehensive review of
the work related to both of the crowd counting and density estimation
using deep network is provided in quite a very unique way to make it
useful for the researchers working in the field of computer vision.
Following are the salient features of our review that play key role in
making our contribution worthwhile and unique among the surveys and
reviews of similar kind: (i) our survey intends to classify the
noteworthy literature with respect to the application and the tasks
related to crowd counting and density estimation, (ii) the survey also
tried to cover in-depth contributions in crowd counting existing works
at different level like images or videos (iii) this state-of-the-art
review covers each article in the following dimensions: the designated
task performed, source level, results obtained, features used, and (iv)
lastly it concludes the summary of the related articles according to the
publishing years, related tasks (or subtasks), and types of classifiers
used. In the end, major challenges and tasks related to crowd counting
and density estimation in computer vision are also discussed.