A Survey of Recent Advances in CNN-based Single Image Crowd Counting and Density Estimation

By Vishwanath A. Sindagi, Vishal M. Patel. 2017

581 citations recorded by the source corpus.

Abstract Estimating count and density maps from crowd images has a wide range of applications such as video surveillance, traffic monitoring, public safety and urban planning. In addition, techniques developed for crowd counting can be applied to related tasks in other fields of study such as cell microscopy, vehicle counting and environmental survey. The task of crowd counting and density map estimation is riddled with many challenges such as occlusions, non-uniform density, intra-scene and inter-scene variations in scale and perspective. Nevertheless, over the last few years, crowd count analysis has evolved from earlier methods that are often limited to small variations in crowd density and scales to the current state-of-the-art methods that have developed the ability to perform successfully on a wide range of scenarios. The success of crowd counting methods in the recent years can be largely attributed to deep learning and publications of challenging datasets. In this paper, we provide a comprehensive survey of recent Convolutional Neural Network (CNN) based approaches that have demonstrated significant improvements over earlier methods that rely largely on hand-crafted representations. First, we briefly review the pioneering methods that use hand-crafted representations and then we delve in detail into the deep learning-based approaches and recently published datasets. Furthermore, we discuss the merits and drawbacks of existing CNN-based approaches and identify promising avenues of research in this rapidly evolving field.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Domain Adaptation for Traffic Density Estimation (2021, methodology): “Following the taxonomy adopted in (Sindagi and Patel, 2018), we can broadly classify existing counting approaches into two categories: counting by detection and counting by regression.”
Scale Aggregation Network for Accurate and Efficient Crowd Counting (2018, background): “[25] introduced a comprehensive survey of CNN-based counting approaches.”
DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density Estimation (2017, methodology): “We follow the convention of existing works [32, 41, 23] to use the mean absolute error (MAE) and mean squared error (MSE) as the evaluation metric.”
Where are the Blobs: Counting by Localization with Point Supervision (2018, methodology): “For datasets with single-class objects, we report the mean absolute error (MAE) which measures the deviation of the predicted count p i from the true count c i , computed as 1 N MAE is a commonly used metric for evaluating object counting methods [39,40].”
Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs (2017, background): “A comprehensive survey of recent cnnbased methods for crowd counting can be found in [33].”
MoVie: Revisiting Modulated Convolutions for Visual Counting and Beyond (2021, background): “However, such models are designed to deal exclusively with a single category of interest, and usually require point-level supervision (Sindagi & Patel, 2017) in addition to the ground-truth overall count number for training.”
Crowd Counting via Adversarial Cross-Scale Consistency Pursuit (2018, background): “A comprehensive survey of recent CNN-based methods for crowd counting can be found in [29].”
Counting Vehicles with Deep Learning in Onboard UAV Imagery (2019, methodology): “Most of the current methods are CNNs-based and basically, we can broadly classify them into two categories [3]: regression-based and detection-based approaches.”
Counting Vehicles with Cameras (2018, methodology): “Following the taxonomies adopted in [11] and [14] we can divide counting approaches into four main categories: counting by detection, counting by clustering, counting by regression and counting by density estimation.”
A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications (2023, background): “The density NN refers to Kernel Density Estimation that applies small groups with closely linked data in order to estimate new points dissemination in problem space [37, 38].”
Crowd Counting and Density Estimation by Trellis Encoder-Decoder Networks (2019, background): “In this section, we provide a brief review of the most related work and refer to comprehensive surveys for crowd counting [35, 36, 41, 13].”
Improving Object Counting with Heatmap Regulation (2018, background): “Sindagi and Patel [25] provide a comprehensive survey of these approaches.”
Learning Multi-Level Density Maps for Crowd Counting (2020, background): “EFFECTIVE crowd counting from images or videos has become an increasingly important task in the field of computer vision [1]–[4].”
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (2018, methodology): “We divide our discussion of related work into two main groups as in [32]: traditional approaches and CNN-based methods.”