Cross-scene crowd counting via deep convolutional neural networks

By Cong Zhang, Hongsheng Li, Xiaogang Wang, Xiaokang Yang. 2015

1,185 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Safety, Risk, Reliability and Quality

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

In Safety, Risk, Reliability and Quality

Citations in the corpus

DeepSetNet: Predicting Sets with Deep Neural Networks (2016, methodology): “It is important to bear in mind that unlike many existing approaches that learn to count [1, 5, 12, 20, 24, 35, 50, 51], our main goal is not object counting.”
Representation Learning by Learning to Count (2017, methodology): “There are methods to count humans in crowds [4, 42, 8], cars [28], and penguins [2].”
Linking Image and Text with 2-Way Nets (2016, background): “In the literature, the Euclidean loss is often combined with other losses [36, 50], 0 0.”
A Survey of Recent Advances in CNN-based Single Image Crowd Counting and Density Estimation (2017, methodology): “In this section, we review ﬁve key datasets [15, 20, 41, 107, 111] followed by a discussion on the results of CNN-based approaches and recent traditional methods that were not included in the earlier surveys.”
Towards Perspective-Free Object Counting with Deep Learning (2016, methodology): “[7], our model gets a better performance in 3 of the 4 subsets.”
DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density Estimation (2017, methodology): “Our proposed method is evaluated on three major crowd counting datasets [6, 41, 42] collected from real-world surveillance cameras.”
Learning Multi-Level Density Maps for Crowd Counting (2020, methodology): “To be consistent with previous works [23], we prune the last convolutional layer so that the features out of ROI regions are set to zero.”
A Large Contextual Dataset for Classification, Detection and Counting of Cars with Deep Learning (2016, background): “Most contemporary counting methods can be broadly categorized as a density estimator [14,15,16] or, detection instance counter [11,13,17].”
Deep Count: Fruit Counting Based on Deep Simulated Learning (2017, neutral): “Experimental Results with Real Data representative images along with their predicted and actual count. for training, the network performs well for images for training, the network performs well real images for training, the network performs having real images for training, the network performs for t”
JHU-CROWD++: Large-Scale Crowd Counting Dataset and A Benchmark Method (2020, methodology): “Existing efforts like UCF CROWD 50 [3], World Expo ’10 [4] and ShanghaiTech [28] have progressively increased the complexity of the datasets in terms of average count per image, image diversity etc.”
Incorporating Side Information by Adaptive Convolution (2017, methodology): “…CSRNet (Li et al. 2018; 3) we conduct more thorough experiments on crowd counting, including results using FCN and AFCN, and results on WorldExpo’10 (Zhang et al. 2015; 4) we perform various ablation experiments to study the helpfulness of different kinds of side information, the effect of the…”
Improving Object Counting with Heatmap Regulation (2018, methodology): “Also, we generate GAM only around the dot annotations for the heads of people in the image which needs less effort compared to the crowd-counting paper [9], where the activation maps are generated using 2 different Gaussians for both head and body considering their resolution after perspective proje”
Switching Convolutional Neural Network for Crowd Counting (2017, methodology): “We follow the approach of other state-of-the-art models [18, 2, 9, 19] and use 5fold cross-validation to validate the performance of SwitchCNN on UCF CC 50.”
Attention Scaling for Crowd Counting (2020, methodology): “And the WorldExpo’10 [46] dataset provides cross-scene test sets that can test the adaptive capacity of the network for different scenes.”
Guided Attention Network for Object Detection and Counting on Drones (2019, methodology): “Object counting methods aim to predict the total number of objects in different categories existing in images, such as pedestrian counting [20, 26, 46, 50], vehicle counting [12, 49], goods counting [11, 21] and general object counting [2, 3, 18].”
Context-Aware Crowd Counting (2018, methodology): “As in previous work [41, 42, 31, 30, 19, 5, 21, 36, 32, 28, 33] on this dataset, we report the MAE of each scene, as well as the average over all scenes.”
Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs (2017, methodology): “[44] and it contains 3,980 annotated frames from 1,132 video sequences captured by 108 surveillance cameras.”
Composition Loss for Counting, Density Map Estimation and Localization in Dense Crowds (2018, methodology): “[29] train a CNN alternatively to predict density map and count in a patch, and then average the density map for all the overlapping patches to obtain density map for the entire image.”