Feature Mining for Localised Crowd Counting

By Ke Chen, Chen Change Loy, Shaogang Gong, Tony Xiang. British Machine Vision Conference, 2012

668 citations recorded by the source corpus.

This paper presents a multi-output regression model for crowd counting in public scenes. Existing counting by regression methods either learn a single model for global counting, or train a large number of separate regressors for localised density estimation. In contrast, our single regression model based approach is able to estimate people count in spatially localised regions and is more scalable without the need for training a large number of regressors proportional to the number of local regions. In particular, the proposed model automatically learns the functional mapping between interdependent low-level features and multi-dimensional structured outputs. The model is able to discover the inherent importance of different features for people counting at different spatial locations. Extensive evaluations on an existing crowd analysis benchmark dataset and a new more challenging dataset demonstrate the effectiveness of our approach.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Transportation

View Paper: https://www.semanticscholar.org/paper/493b0e0977370c0eae566b558cbcaea4aadf6634

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 Transportation

Citations in the corpus

Multi-source Multi-scale Counting in Extremely Dense Crowd Images (2013, methodology): “, USCD dataset with density of 11 − 46 people per frame [4], Mall dataset with density of 13−53 individuals per frame [5], and PETS dataset containing 3 − 40 people per frame [9].”
Cumulative Attribute Space for Age and Crowd Density Estimation (2013, methodology): “This can be Active Appearance Model features [9] for age estimation and foreground & edges & GLCM features [4, 8] for crowd counting.”
Extremely Overlapping Vehicle Counting (2015, background): “We understand the extremely overlapping vehicle counting problem as a variant of the problem of crowd counting [7,8,9,13,14,15].”
Switching Convolutional Neural Network for Crowd Counting (2017, neutral): “Cumulative Attribute Regression [5] 2.”
Distribution Matching for Crowd Counting (2020, methodology): “Later methods [5, 6, 49, 7] avoid the detection problem and directly learn to regress the count from a feature vector.”
DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density Estimation (2017, methodology): “The Mall dataset [6] contains 2000 frames, collected in a shopping mall.”
Spatiotemporal Group Context for Pedestrian Counting (2014, background): “Obviously, most of the detection-based approaches rely on the training process and most of them are based on a single frame, so they have difficulties to achieve stable results due to the view changes, scale changes, pose changes, and spatial overlapping.”
Scale Aggregation Network for Accurate and Efficient Crowd Counting (2018, methodology): “Researchers attempted regression-based methods to directly learn a mapping from the feature of image patches to the count in the region [18,19,20].”
Learning From Synthetic Data for Crowd Counting in the Wild (2019, background): “Dataset Number Average Count Statistics of Images Resolution Total Min Ave Max UCSD [8] 2,000 158× 238 49,885 11 25 46 Mall [9] 2,000 480× 640 62,325 13 31 53 UCF CC 50 [14] 50 2101× 2888 63,974 94 1,279 4,543 WorldExpo’10[41] 3,980 576× 720 199,923 1 50 253 SHT A [43] 482 589× 868 241,677 33 501 3,”
Where are the Blobs: Counting by Localization with Point Supervision (2018, methodology): “Our benchmark uses datasets representing diﬀerent settings for object counting: Mall [15], UCSD [16], and Shang-haiTech B [17] as crowd datasets; MIT Traﬃc [18], and Park lot [5] as surveillance datasets; Trancos [6] as a traﬃc monitoring dataset; and Penguins [1] as a population monitoring dataset.”
Drone-Based Object Counting by Spatially Regularized Regional Proposal Network (2017, methodology): “Several methods [3, 6, 7, 8, 15] try to predict counts by using global regressors trained with low-level features.”
Context-Aware Crowd Counting (2018, background): “Even though approaches relying on low-level features [9, 6, 4, 27, 7, 14] can yield good results, they have now mostly been superseded by CNN-based methods [42, 31, 5], a survey of which can be found in [36].”
Composition Loss for Counting, Density Map Estimation and Localization in Dense Crowds (2018, methodology): “Earlier regression-based approaches mapped global image features or a combination of local patch features to obtain counts [15], [5], [12], [6].”
Understanding Traffic Density from Large-Scale Web Camera Data (2017, neutral): “Ridge Regression [6] 2.”
Crowd Counting via Adversarial Cross-Scale Consistency Pursuit (2018, methodology): “2.16 7.45 Cumulative Attributes [6] 2.07 6.86 Zhang et al .”
Crowded Scene Analysis: A Survey (2015, result): “Differently from previous survey such as [15], crowd counting or density estimation, a closely related topic [35]–[37], is not covered in this survey.”