Search

Crowd counting and density estimation are the principal objectives of crowd analysis, which offer significant applications in surveillance, event management, and traffic design. In the field of crowd flow, including simulations, the dynamics of crowd movement exhibit characteristics such as followability and, thus, are categorized under a distinct flow paradigm. The recent advancements in deep learning have propelled the usage of neural networks tailored for crowd counting and density estimation from video feeds. Nonetheless, prior models did not consider crowd dynamics. This study proposes a novel method that combines neural networks with crowd dynamics. Specifically, we introduced a new penalty term that represents prior knowledge of crowd dynamics and refined the neural network outputs via static/dynamic floor field models, and grid-based crowd dynamics models. Empirical evaluation on benchmark datasets demonstrated the superiority of the proposed method over existing state-of-the-art techniques. Further analysis of each scene confirmed that the crowd counting performance is highly dependent on the scene, and the impact of the three methodological components (i.e., the penalty term and the two-floor fields) on performance varies across scenes. In particular, the floor-field model tended to be more effective when there were no significant changes in the scene. Our code is available on GitHub. https://github.com/hanebarla/ FF-guided-NeuralCC

In this paper, we propose a novel and enhanced approach for crowd counting within the domain of manatee monitoring, aiming to significantly improve efficiency and accuracy. The proposed model achieves state-of-the-art results in the challenging task of manatee counting, simplifying the work of scientists and experts in the field. Our model not only facilitates the identification and enumeration of manatees in images and videos but also excels in scenarios that pose considerable challenges for human observers. To enhance accurate counting of the manatee aggregation, we introduce a framework with three key innovations to tackle the challenge: a new approach to generate density maps during the training process, an augmented technique to balance the dataset, and a cross-domain solution to enhance overall performance. The proposed two-dimensional Gaussian kernel offers a refined method for creating density maps, providing a more robust foundation for the training phase. Additionally, we built a balanced and augmented dataset, ensuring that the model is exposed to diverse and representative instances, thus improving its generalization capabilities. Furthermore, we incorporate a cross-domain phase pretraining the model utilizing an image dataset of wild animals to initialize the weights and further improve performance. Experiments and comparisons, with respect to previously established CSRNET model presented in Wang et al. (2023), demonstrate noteworthy improvements. Remarkably, our model achieves a Mean Absolute Error (MAE) of nearly half compared to the rival approach, showcasing the substantial advancements achieved through our refined methodology. This progress boosts the reliability of manatee counting in conservation efforts and ecological research.

Recent work in crowd counting focuses on counting over detected individuals rather than estimating the number of people in the image. However, existing crowd localization methods directly detect the head point or region of individuals, which may entail non-responsibility of the outputs that fall outside the grid. Our proposed Pseudo Square Label Network (PSL-Net) presents a novel method for crowd counting and localization, which takes advantage of the anchor-free detection in which PSL-Net predicts the probability of the center point that fall into the responsible grid, while indirectly detecting an individual outside of the responsible grid through box regression and centerness estimation. This study proposes to supervise with pseudo square label(PSL), which is generated around point annotation with fixed size. Furthermore, we design a partial many-to-one matching algorithm to assign precise labels by only matching within PSL during the training phase, and associate the predicted points with their responsible grids through centerness during the inference phase. As a result, not only PSL-Net achieves state-of-the-art on ShanghaiTech Part A and B, which are the most popular datasets in crowd counting, but also achieves state-of-the-art among the point detection-based methods in crowd localization.

Accurately measuring the level of crowding in transit cars is crucial for ensuring passenger safety and efficient operation. However, applying object detection algorithms to crowd counting in transit cars poses difficulties due to the low viewpoint of the cameras and the labor-intensive task of image labeling. Although some researchers have explored regression-based crowd counting methods without labeling with bounding boxes, their approaches still necessitate manual counting of passengers for image labeling. To overcome these challenges, we propose a novel calibration method for regression-based models that minimizes the number of labeled images required for training. Our approach employs image pairs and triplets with ranks for reinforcing the model training. Subsequently, the training task requires a minimal number of images labeled with exact passenger counts. Experimental results demonstrate that our proposed calibration approach considerably enhances the crowd counting performance of the conventional regression-based model. Specifically, our method reduces the mean absolute error (MAE) by 76.5% and 34.3% for conventional detection- and regression-based calibration methods, respectively.

With the rapid development of society, crowded scenes can be seen almost everywhere. Therefore, it is important to accurately predict the number and density distribution of people in those crowded regions by utilizing image analysis. In recent years, most of those studies based on various deep learning technologies for pedestrian image analysis have been based on mature convolution neural networks (CNN). Nowadays, vision transformer has demonstrated its competitive performance compared with CNNs in many computer vision domains, and provides a novel idea for density distribution in image. In this paper, we modify the Swin-Transformer and integrate CNN to propose a feature aggregation and stitching network (FASNet), which effectively improves the counting accuracy. The hierarchical vision transformer backbone captures the global multi-scale features of the image, and encode the interaction information among different pedestrians in the deep network. Feature Aggregation Module (FAM) is used to fuse the deep and shallow features, and then Density Regression Module (DRM) upsamples the output of FAM and finally produces the predicted crowd density map and the final count. In addition, we propose the Feature Stitching Mechanism (FSM) to cope with the feature damage or loss caused by image cropping during the model testing. The experimental results on three benchmark datasets (UCF_CC_50, UCF_QNRF, Shanghaitech) demonstrate the effectiveness of our proposed scheme.

Crowd analysis in general and counting in congested scenes, in particular, is an effective and vibrant research domain in computer vision due to its numerous applications. Understanding the risk analysis and safety aspects of crowd dynamics at various vital occasions related to sports cultural and religious activities, specifically, at Hajj and Umrah, is essential. Thousands of people gathered in a small area to carry out their rites. Localizing and counting the annotated head points is quite challenging due to occlusion and large-scale variation in the congested environment. To deal with these problems, a small and effective solution is to generate the density maps. However, the significant flaws of the density map have a blurry Gaussian blob which is less effective for counting and localizing head annotations in the congested scene. To overcome these issues, we propose Congested Scene Crowd Counting and Localization Network (CSCCL-Net) with a Focal inverse Distance Transform (FIDT) map that can count and localize the people simultaneously in the highly congested scene. To evaluate the proposed model’s efficiency, extensive tests were performed on the ShanghaiTech part A, ShanghaiTech part B, and JHU-CROWD++ datasets. The proposed model outperforms existing state-of-the-art techniques regarding high accuracy and low Mean Absolute Error (MAE) and Mean Square Error (MSE) values.

Due to extreme scale variations in highly congested scenes, the accuracy of CNN-based crowd counting approaches still has considerable room for further improvements. In this paper, we propose a new multi-scale feature adaptive integrated network (MSFAINet) for crowd counting that adopts the multi-scale feature, hybrid attention, and dilated convolution. First, the proposed MSFAINet extracts feature maps from different levels by the improved VGG16 and focuses on more important information that represents features at different scales. Second, it adopts a hybrid attention mechanism to enhance the receptive field of an image while reducing the loss of feature information caused by channel competition and then passes these features into the dilated convolution combined with the traditional convolution. Finally, it generates the density estimation map by accelerating the convergence of the network. The proposed MSFAINet is used to conduct extensive studies to demonstrate the effectiveness of the approach on several mainstream datasets. From the experimental results, MSFAINet can extract and retain more detailed information and greatly reduce the influence of scale variations in crowd counting.

Deep learning-based crowd density estimation can greatly improve the accuracy of crowd counting. Though a Bayesian loss method resolves the two problems of the need of a hand-crafted ground truth (GT) density and noisy annotations, counting accurately in high-congested scenes remains a challenging issue. In a crowd scene, people’s appearances change according to the scale of each individual (i.e., the person-scale). Also, the lower the sparsity of a local region (i.e., the crowd-sparsity), the more difficult it is to estimate the crowd density. In this paper, we propose a novel congestion-aware Bayesian loss method that considers the person-scale and crowd-sparsity. We estimate the person-scale based on scene geometry, and we then estimate the crowd-sparsity using the estimated person-scale. The estimated person-scale and crowd-sparsity are utilized in the novel congestion-aware Bayesian loss method to improve the supervising representation of the point annotations. We verified the effectiveness of each proposed component through several ablation experiments, and in the various experiments on public datasets, our proposed method achieved state-of-the-art performance.

Counting the number of people in a crowd has gained attention in the last decade. Due to its benefit to many applications such as crowd behavior analysis, crowd management, and video surveillance systems, etc. Counting crowded scenes, like stadiums, represents a challenging task due to the inherent occlusions and density of the crowd inside and outside the stadiums. Finding a pattern to control thousands of people and counting them is a challenging task. With the introduction of Convolutional Neural Networks (CNN), enables performing this task with acceptable performance. The accuracy of a CNN-based method is related to the size of data used for training. The availability of the dataset is sparse. In particular, there is no dataset in the literature that can be used for training applications for crowd scene. This paper proposes two main contributions including a new dataset for crowd counting, and a CNN-based method for counting the number of people and generating the crowd density maps. The proposed dataset for Football Supporters Crowd (FSC-Set) is composed of 6000 annotated images (manually) of different types of scenes that contain thousands of people gathering in or around the stadiums. FSC-Set contains more than 1.5 Million individuals. The collected images are captured under varying Fields of Views (FOV), illuminations, resolutions, and scales. The proposed dataset can also be utilized for other applications, such as individual’s localization and face detection as well as team recognition from supporter images. Further, we propose a CNN-based method named FSCNet for crowd counting exploiting context-aware attention, spatial-wise attention, and channel-wise attention modules. The proposed method is evaluated on our established FSC-Set and other existing datasets then compared to state-of-the-art methods. The obtained results show satisfactory performances on all the datasets. The dataset is made publicly available and can be requested using the following link: https://sites.google.com/view/fscrowd-dataset/

Wireless sensing represented by WiFi channel state information (CSI) is now enabling various fields of applications such as person identification, human activity recognition, occupancy detection, localization, and crowd estimation these days. So far, those fields are mostly considered as separate topics in WiFi CSI-based methods, on the contrary, some camera and vision-based crowd estimation systems intuitively estimate both crowd size and location at the same time. Our work is inspired by the idea that WiFi CSI also may be able to perform the same as the camera does. In this paper, we construct Wi-CaL, a simultaneous crowd counting and localization system by using ESP32 modules for WiFi links. We extract several features that contribute to dynamic state (moving crowd) and static state (location of the crowd) from the CSI bundles, then assess our system by both conventional machine learning (ML) and deep learning (DL). As a result of ML-based evaluation, we achieved 0.35 median absolute error (MAE) of counting and 91.4% of localization accuracy with five people in a small-sized room, and 0.41 MAE of counting and 98.1% of localization accuracy with 10 people in a medium-sized room, by leave-one-session-out cross-validation. We compared our result with percentage of non-zero elements metric (PEM), which is a state-of-the-art metric for crowd counting, and confirmed that our system shows higher performance (0.41 MAE, 81.8% of within-1-person error) than PEM (0.62 MAE, 66.5% of within-1-person error).

Crowd density estimation is an important topic in computer vision due to its widespread applications in surveillance, urban planning, and intelligence gathering. Resulting from extensive analysis, crowd density estimation reflects many aspects such as similarity of appearance between people, background components, and inter-blocking in intense crowds. In this paper, we are interested to apply machine learning for crowd management in order to monitor populated area and prevent congestion situations. We propose a Single-Convolutional Neural Network with Three Layers (S-CNN3) model to count the number of people in a scene and conclude about the crowd estimation. Then, a comparative study for density counting establishes the performance of the proposed model against the convolutional neural networks with four layers (single-CNN4) and Switched Convolutional neural networks (SCNN). ShanghaiTech dataset, considered as the largest data base for crowd counting, is used in this work. The proposed model proves high effectiveness and efficiency for crowd density estimation with 99.88% of average test accuracy and 0.02 of average validation loss. These results achieve better performance than the existing state-of-the-art models.

We propose a strategy that focuses on estimating the number of people in a crowd, one of the aims of crowd analysis, using static images or video images. While manual feature extraction was not performed with pixel and regression-based methods in the first studies on crowd analysis, recent studies use Convolutional Neural Networks (CNN) based models. However, it is still difficult to extract spatial information such as position, orientation, posture, and angular value for crowd estimation from a density map. This study uses capsule networks and routing by agreement algorithm as an attention module. Our proposed approach consists of both CNN and capsule network-based attention modules in a two-column deep neural network architecture. We evaluate our proposed approach compared with other state-of-the-art methods using three well-known datasets: UCF-QNRF, UCF_CC_50, UCSD, ShangaiTech Part A, and WorldExpo’10.

Crowd counting considers one of the most significant and challenging issues in computer vision and deep learning communities, whose applications are being utilized for various tasks. While this issue is well studied, it remains an open challenge to manage perspective distortions and scale variations. How well these problems are resolved has a huge impact on predicting a high-quality crowd density map. In this study, a hybrid and modified deep neural network (U-ASD Net), based on U-Net and adaptive scenario discovery (ASD), is proposed to get precise and effective crowd counting. The U part is produced by replacing the nearest upsampling in the encoder of U-Net with max-unpooling. This modification provides a better crowd counting performance by capturing more spatial information. The max-unpooling layers upsample the feature maps based on the max locations held from the downsampling process. The ASD part is constructed with three light pathways, two of which have been learned to reflect various densities of the crowd and define the appropriate geometric configuration employing various sizes of the receptive field. The third pathway is an adaptation path, which implicitly discovers and models complex scenarios to recalibrate pathway-wise responses adaptively. ASD has no additional branches to avoid increasing the complexity. The designed model is end-to-end trainable. This integration provides an effective model to count crowds in both dense and sparse datasets. It also predicts an elevated quality density map with a high structural similarity index and a high peak signal-to-noise ratio. Several comprehensive experiments on four popular datasets for crowd counting have been carried out to demonstrate the proposed method’s promising performance compared to other state-of-the-art approaches. Furthermore, a new dataset with its manual annotations, called Haramain with three different scenes and different densities, is introduced and used for evaluating the U-ASD Net.

Crowd counting is considered as the essential computer vision application that uses the convolutional neural network to model the crowd density as the regression task. However, the vision-based models are hard to extract the feature under low-quality conditions. As we know, visual and audio are used widely as media platforms for human beings to touch the physical change of the world. The cross-modal information gives us an alternative method of solving the crowd counting task. In this case, in order to solve this problem, a model named the Audio-Visual Multi-Scale Network (AVMSN) is established to model the unconstrained visual and audio sources for completing the crowd counting task in this paper. Based on the Feature extraction and Multi-modal fusion module, in order to handle the objects of various sizes in the crowd scene, the Sample Convolutional Blocks are adopted by the AVMSN as the multi-scale Vision-end branch in the Feature extraction module to calculate the weighted-visual feature. Besides, the audio, which is the temporal domain transformed into the spectrogram information and the audio feature is learned by the audio-VGG network. Finally, the weighted-visual and audio features are fused by the Multi-modal fusion module, which adopts the cascade fusion architecture to calculate the estimated density map. The experimental results show the proposed AVMSN achieves a lower mean absolute error than other state-of-art crowd counting models under the low-quality conditions.

Density estimation aims to predict the spatial distribution of a crowd scene, and crowd counting aims to automatically check the number of heads as close as the ground truth. We propose a mask guided GAN (Generative Adversarial Network) architecture to solve these two problems synthetically. Step one is generating a segmentation mask, separating the crowd region from the background and redundant information. Step two is predicting the density map with an adversarial learning process guided by the former mask information. Moreover, we branch out from the base network and feed into a counting regressor, which is solely to provide more accurate counting results. The whole scheme is trained collaboratively by compositing density loss (a weighted loss to balance the influence of different data) and counting loss (an MSE loss for counting regression branch). Through the mask information guidance, GAN likely finds its way training to capture more distinguishing features. Therefore, we also achieve more accurate prediction results. Experimental results on different datasets including collected from the internet and the actual scene in Shanghai, the second-most populous city located on China's southeast coast, indicate the validity and robustness by comparable counting numbers and high-quality density maps focusing on crowd area.

The main challenge of crowd counting is the dramatic variations in scale and perspective. Most methods combine features of different receptive fields to deal with scale problems, but do not perform well on continuous scale variations in local areas. Moreover, they use only Euclidean loss which assumes independence of each pixel without considering local correlation in the density map. Therefore, we propose the enhanced scale robust network (ESRN) for accurate and efficient crowd counting. ESRN contains an embedded GAN module, which is followed by a well-designed enhancer. The discriminator guides the generator to strengthen the local correlation and create high-quality density maps, while the enhancer combines context information in different scales and varying among different sub-regions to further enhance the scale robustness of the network. The embedded GAN module is jointly trained with the enhancer. Extensive experimental results on three challenging datasets well demonstrate the effectiveness of our method.

In this paper we propose a novel regression based RGBD crowd counting method. Compared with previous RGBD crowd counting methods which mainly exploit depth cue to facilitate person/head detection, our approach adopts density map regression and is more robust to severe occlusion under dense crowded scenarios. We develop a cascaded depth-aware counting network that jointly performs head segmentation and density map regression. Our network explicitly feeds depth map at each stage so that depth cues are sufficiently exploited. The multi-task strategy allows the network to explicitly attent to foreground regions of a crowd scene and improve density regression. To generate the ground truth of head segmentation and density map, we propose a head scale estimation method according to the basic geometric assumption and camera projection function. Experiments on two public RGBD crowd counting benchmarks, ShanghaiTechRGBD dataset and MICC dataset show that the proposed method achieves new state-of-the-art on both datasets. Further, our method can be easily extended to RGB datasets and achieves comparable performances on WorldExpo'10 dataset and UCF-QNRF dataset.

Crowd counting has gained popularity due to wide applications, such as intelligent security, and urban planning. However, scale variation and perspective distortion make it a challenging task. Most existing works focus on multi-scale feature extraction to address the challenge of scale variation and perspective distortion. In this paper, we propose a novel Global Context Instructive Network (GCINet), which devotes to making full use of extracted features and obtaining precise counts. The main contributions are four folds. First, we construct a three-column Feature Processor to generate features with different scales. Second, an Instructive Module is proposed to introduce global context which is the substance for generating adaptive features. Based on global context, the three-column Feature Processor constitutes an adaptive feature generator. Third, a novel loss function which integrates Euclidean distance and spatial correlation is proposed to enhance the spatial correlation and consistency between pixels. We no longer regard a pixel as an independent point in the calculation, but consider the neighborhood space of the pixel to achieve complementary effects. Finally, we conduct experiments on ShanghaiTech dataset, UCF_CC_50 dataset, UCF_QNRF dataset and UCSD dataset which show that our approach achieves state-of-the-art performance.

Pedestrian head detection plays an important role in identifying and localizing individuals in real world visual data. Head detection is a nontrivial problem due to considerable variance in camera view-points, scales, human poses, and appearances in the scene. Thanks to the translation invariance property of convolutional neural networks (CNNs) which enables large capacity CNNs to handle the problem of appearance and pose variations in the scene. However, the problem of scale invariance is still an open issue. To address this problem, this paper presents a two-stage head detection framework that utilizes fully convolutional network (FCN) to generate scale-aware proposals followed by CNN that classifies each proposal into two classes, i.e. head and background. Experiments results show that using scale-aware proposals obtained by FCN, the object recall rate and mean average precision (mAP) are improved. Additionaly, we demonstrate that our framework achieved state-of-the-art results on four challenging benchmark datasets, i.e. HollywoodHeads, Casablanca, SHOCK, and WIDERFACE.

In recent years, deep learning based crowd counting networks have achieved significant progress. However, most of them generate rough crowd density maps due to low-resolution features used for estimating crowd distribution, which affects the performance of crowd counting. To solve this problem, in this paper, we propose a Hierarchical Attention Guided Network (HAGN) for crowd counting. We apply the first 13 layers of VGG-16 to extract base features. Then, the extracted features are processed by the Hierarchical Attention Mechanism (HAM), which guided the extracted features to enlarge step by step via our proposed attention guided branch. Finally, the outputs of HAM are fed to 1 × 1 convolutional layer for final crowd density estimation. Experiments are performed on ShanghaiTech and UCF-QNRF datasets, and our HAGN achieves promising performance compared with the other state-of-the-art methods on crowd counting and crowd localization, respectively.

The major challenges for density maps estimation and accurate counting stem from the large-scale variations, serious occlusions, and perspective distortions. Existing methods generally suffer from the blurred density maps, which are caused by average convolution kernel, and the ineffective estimation across different crowd scenes. In this paper, we propose a multi-scale fusion conditional generative adversarial network (MFC-GAN) that can generate high-resolution and high-quality density maps. The fusion module of MFC-GAN is embedded in a multi-scale generator and discriminator architecture with a novel adversarial loss, which is designed to guide high-resolution density maps generation. In order to address the problem of scale variation, we further propose a bidirectional fusion module. It combines deep global semantic features and shallow local information by leveraging feature maps presented in different layers of the generator. Furthermore, in order to increase the effectiveness of the multi-scale fusion, we design a cross-attention fusion module, which weights the multi-scale fused feature and learns context-aware feature maps for generating high quality density maps. The experiments on four challenging datasets show the effectiveness, feasibility and robustness of the proposed MFC-GAN.

Nowadays, crowd analysis is one of the most important concepts that needs be relied upon, it contributes to decision making and ensuring the safety and security of the crowd. There are a variety of interesting research problems within the scope of crowd analysis including crowd tracking, crowd behaviour recognition and crowd counting. Crowd counting based on images and videos has been studied in past years. Nonetheless, estimating and detecting the number of human heads remains a challenging task due to occlusions, resolution, and lighting changes. This paper provides an overview and performance comparison of crowd counting techniques using convolutional neural networks (CNN) based on density map estimation. In this paper, we present a comprehensive analysis and benchmarking of crowd counting based on the UCF-QNRF dataset that contains the largest number of crowd count images and head annotations available in the public domain. We also show the density maps generation and their empirical evaluation along with performance comparison.

CNN-based crowd counting methods have achieved great progress in recent years. However, most of these CNN-based crowd counting methods do not make full use of contextual information, which contains high-level semantic features and low-level detail features from different receptive fields of CNN. But rich contextual information is important to solve the scale variation problem of crowd counting. So the precision of previous CNN-based crowd counting methods is decreased. To solve this problem, we propose an adaptive attention fusion mechanism (AAFM). AAFM can use multi-scale features from different receptive fields of CNN effectively. It integrates the convolution network for feature learning and the attention mechanism for multi-scale features fusion. We apply the first 13 convolution layers of VGG-16 as the encoder module to extract the base features. Then, the base features are fed into the decoder module. The decoder module mainly contains Density Regression Branch (DRB) and Feature Fusion Branch (FFB). DRB uses multiple convolution layers for feature learning and multi-scale feature extraction. FFB uses attention modules for modeling multi-scale features and element-wise multiply for features fusion. Therefore, AAFM can obtain rich contextual information into the encoder-decoder framework for generating high-quality crowd density maps and accurate counting. We perform experiments on ShanghaiTech, UCF-CC-50, and UCF-QNRF datasets, and AAFM achieves promising results.

Crowd counting is a task that aims to estimate the number of people in an image. Recent crowd counting methods make significant progress by employing convolutional neural networks to regress crowd density maps. One of the most challenging problems in this task is the drastic scale variation of the region of interest in images. In this paper, a Feature Fusion Attention Network (FFANet) is proposed for crowd counting. Firstly, the VGG16 network is adapted as the backbone of the FFANet to extract the features of crowd images. Then, the extracted features are fused by the subsequent two stages. Specifically, the information enhancement operations on the multi-levels features are conducted by Feature Fusion Attention Module (FFAM), which are further refined by the Residual Block (RB). Finally, the features are processed by the Compression Module (CM) to generate a density map. To demonstrate the effectiveness, the proposed algorithm is verified on three benchmark datasets. Evaluation of the algorithm performances in comparison with other state-of-the-art methods indicates the proposed FFANet outperforms the existing methods.

Crowd counting is a challenging task due to the influence of various factors, such as scene transformation, complex crowd distribution, uneven illumination, and occlusion. To overcome such problems, scale-adaptive convolutional neural network (SaCNN) used a convolutional neural network to obtain high-quality crowd density map estimation and integrate the density map to get the estimated headcount. To obtain better performance on crowd counting, an improved crowd counting method based on SaCNN was proposed in this paper. The spread parameter, i.e., the standard variance, of geometry-adaptive Gaussian kernel used in SaCNN was optimized to generate a higher quality ground truth density map for training. The absolute count loss with weight 4e-5 was used to jointly optimize with the density map loss to improve the network generalization ability for crowd scenes with few pedestrians. Also, a random cropping method was applied to improve the diversity of training samples to enhance network generalization ability. The experimental results upon ShanghaiTech public dataset showed that the proposed method can obtain more accurate and more robust results on crowd counting than those of SaCNN.

Crowd counting plays an important role in urban management and public security. Recently, deep learning has shown a great advantage in making the quality of crowd counting more accurate. However, how to apply deep learning models to embedded terminals is still a challenging issue. The main contradiction of the problem lies in high demand for computing resources for deep learning and strict limitation of power consumption from embedded devices. In order to achieve the crowd counting by edge computing (in embedded terminals), we propose a tiny model based on convolutional neural networks. The model can be switched into other two forms to adapt tradeoff between the accuracy and efficiency. Especially, different forms share main parameters so as to save storage and avoid retraining. What is more, this model supports variant sizes of input images, which benefits the applications to a variety of embedded devices. The experiments on two different embedded terminals have shown the effectiveness, efficiency, and flexibility of our proposed model.

This paper proposes an automatic scale-adaptive approach with attention mechanism-based crowd spatial information addressing the crowd counting task, i.e. a novel cascaded crowd counting network. The proposed network is composed of a classification sub-network to estimate crowd scales and the main network to predict the corresponding density maps. First, the image serves as the input of the classification network and the main network. Second, according to the estimated crowd scale results, the main network structure is adjusted; simultaneously, the feature in the intermediate layer of the classification network is added stepwise into the main crowd counting network. Then, the semantic feature of the classification network is converted into the crowd spatial information mask via the proposed spatial attention conversion module, and the crowd spatial information mask is weighted into the specific feature of the main network. Last, the crowd density map and the crowd counting result are obtained. The experiments on challenging Mall, the Shanghaitech_A and Shanghaitech_B datasets prove the effectiveness, feasibility, and robustness of the proposed method, and the ablation study demonstrates the structure rationality of the proposed network.

Crowd counting is an important research topic in the field of computer vision. The multi-column convolution neural network (MCNN) has been used in this field and achieved competitive performance. However, when the crowd distribution is uneven, the accuracy of crowd counting based on the MCNN still needs to be improved. In order to adapt to uneven crowd distributions, crowd global density feature is taken into account in this paper. The global density features are extracted and added to the MCNN through the cascaded learning method. Because some detailed features during the down-sampling process will be lost in the MCNN and it will affect the accuracy of the density map, an improved MCNN structure is proposed. In this paper, the max pooling is replaced by max-ave pooling to keep more detailed features and the deconvolutional layers are added to restore the lost details in the down-sampling process. The experimental results in the UCF_CC_50 dataset and the ShanghaiTech dataset show that the proposed method has higher accuracy and stability.

Counting and localization of people in videos consisting of low density to high density crowds encounter many key challenges including complex backgrounds, scale variations, nonuniform distributions, and occlusions. For this purpose, we propose a scale driven convolutional neural network (SD-CNN) model, which is based on the assumption that heads are the dominant and visible features regardless of the density of crowds. To deal with the problem of different scales of heads in different regions of the videos, we annotate a set of heads in random locations of the videos to develop a scale map representing the mapping of head sizes. We then extract scale aware proposals based on the scale map which are fed to the SD-CNN model acting as a head detector. Our model provides a response matrix rendering accurate head positions via nonmaximal suppression. For experimental evaluations, we consider three standard datasets presenting low density to high density crowd scenes. Our proposed SD-CNN model outperforms the state-of-the-art methods in terms of both frame-level and pixel-level analyses.

The rising need of crowd monitoring in public spaces, especially for safety purposes, pushes the research community to propose and experiment novel methods of crowd density estimation. This paper focuses on device-free RF sensing, which does not require the monitored people to carry any electronic device. In particular, the paper proposes and assesses the performance of a crowd density estimation system based on the analysis of the variations of Channel State Information (CSI) computed from unencrypted synchronization signals transmitted by a eNodeB and reflected/scattered by people located in the monitored area. The proposed method uses features extracted from the list of singular values of the CSI secant set. This approach allows to reduce the impact of CSI variations due to noise or HW instability, improving the sensitivity to CSI variations caused by human presence. The average accuracy achieved by the proposed approach is 84%, which is comparable with the accuracy achieved with WiFi based crowd density estimation systems.

The accuracy of object-based computer vision techniques declines due to major challenges originating from large scale variation, varying shape, perspective variation, and lack of side information. To handle these challenges most of the crowd counting methods use multi-columns (restrict themselves to a set of specific density scenes), deploying a deeper and multi-networks for density estimation. However, these techniques suffer a lot of drawbacks such as extraction of identical features from multi-column, computationally complex architecture, overestimate the density estimation in sparse areas, underestimating in dense areas and averaging of feature maps result in reduced quality of density map. To overcome these drawbacks and to provide a state-of-the-art counting accuracy with comparable computational cost, we therefore propose a deeper and wider network: a Context-aware Scale Aggregation CNN-based Crowd Counting method (CASA-Crowd) to obtain the deep, varying scale and perspective varying features. Further, we include a dilated convolution with varying filter size to obtain contextual information. In addition, due to different dilation rates, a variation in receptive field size is more useful to overcome the perspective distortion. The quality of density map is enhanced while preserving the spatial dimension by obtaining a comparable computational complexity. We further evaluate our method on three well-known datasets: UCF_CC_50, ShanghaiTech Part_A, ShanghaiTech Part_B.

Density-based crowd counting methods with deep convolutional neural network (CNN) have achieved the state of the art on the challenging datasets. Experimental results showed that the performance of these methods suffers from two problems: 1) Background interference problem: there are some estimated spurious density values in the background regions which degrade the counting accuracy. 2) Cross-scale problem: the scale of human heads varies greatly in crowd images which lead to poorer quality of the density maps. In this study, we aim to address the two problems for enhancing the counting accuracy. To address the former, a light semantic mask module (SMM) is proposed to learn the semantic masks of crowd images where the ground-truth semantic masks generated from the ground-truth density map are taken as the supervision information. To tackle the latter, we propose a span architecture (SA) to effectively capture the large-scale-variation information in the crowd images by building the cross-scale features from the pyramidal structure of a deep CNN. To adaptively leverage the salient cross-scale features, a Cross-scale Adaptive Module (CAM) is delicately designed. In the end, integrating all elements above, an end-to-end trainable and single-column crowd counting model called the SMCA-CNN is developed and trained with a joint loss function consisting of the cross-entropy loss and Euclidean loss. Extensive experiments on five challenging datasets demonstrate the effectiveness of our SMCA-CNN. Compared with the previous state-of-the-art methods, our model achieves 17.1% lower MAE on dataset UCF_CC_50 and 23.6% lower MAE on the newly published dataset UCF-QNRF.

Cross-scene counting is difficult if only limited training samples are available in the new scene. In this paper, a cross-scene counting model is learned with information transferred from other scenes. Counting is achieved through regression, which maps the features of crowds to their counts. Hand-crafted features are extracted from segmented crowd foregrounds obtained through block robust principal component analysis. Samples of existing scenes (source domain) are adaptively transferred into the new scene (target domain) through domain adaptation. Then, a counting model based on domain adaptation-extreme learning machine (DA-ELM) is efficiently learned via iterative optimization with training samples of both domains. Quantitative analysis indicates that the DA-ELM can count the crowds of a new scene with only a half of the training samples compared with counting without domain adaptation. Contrastive evaluations based on three benchmarking data sets are implemented with several state-of-the-art domain adaptation approaches, including hand-crafted feature-based and deep neural network-based approaches. Results reveal the effectiveness of DA-ELM in transferring information through embedding domain adaptation into an ELM framework.

The major challenges for accurate crowd counting stem from the large variations in the scale, shape, and perspective. In fact, dealing with such difficulties depends on the geometric transformation capabilities of the network. Thus, we propose the deformation aggregation network (DA-Net) that can incrementally incorporate adaptive receptive fields to capture the fine-grained density distribution. Our model starts from a fundamental diamond network based on the successive steps of rising and falling on the channels, which utilizes small-fixed kernels accounting for preserving spatial information. For sake of accommodating to scale variations, we further apply deformable convolutions to deliver precise localization by augmenting the spatial sampling locations with additional offsets, capable of transformation-invariant speciality on diverse scenes. In addition, we assign adjustable weights to supervise network to learn the fusion proportion adaptively and aggregate multiple levels of abstractions to get the final density map. DA-Net is a fully convolutional network that is able to take input of arbitrary size. The proposed method delivers the state-of-the-art performance on four benchmarks (ShanghaiTech, WorldExpo'10, UCSD, and UCF_CC_50). To further verify the generality of our model, we conduct an extended experiment on vehicle data set TRANCOS and achieve significant results.