Marked point processes for crowd counting

Ge, Weina; Collins, Robert T.

doi:10.1109/cvpr.2009.5206621

Cited by 230 publications

(88 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition to the above categories, there have been many recent interesting work in counting using Bayesian modeling. Pham et al [33] employs point process inference for large scale object detection and counting, while in [13], a Bayesian marked point process is developed to detect and count people in crowded scenes, leading to an estimate of the count, location and pose of each person in the scene. Point processes allow convenient modeling and analysis of spatial data, the object configuration and the interaction between objects.…”

Section: Counting By Detectionmentioning

confidence: 99%

Cox Processes for Counting by Detection

Rajan

Jedynak

2018

J Math Imaging Vis

View full text Add to dashboard Cite

In this work, Cox processes and Convolutional Neural Net classifiers (CNNs) are used to estimate the number of instances of an object in an image. Poisson processes are well suited to model events that occur randomly in space, such as the location of objects in an image or the enumeration of objects in a scene. Doubly Stochastic Poisson (or Cox) processes offer increased flexibility, but their computational complexity and storage requirements do not easily scale with image size, typically requiring O(n 3) computation time and O(n 2) storage, where n is the number of pixels in an image. To mitigate this problem we employ the Kronecker algebra, which takes advantage of direct product structures. As the likelihood is non-Gaussian, the Laplace approximation is used for inference, employing the conjugate gradient and Newton's method. Our approach has then close to linear performance, requiring only O(n 3/2) computation time and O(n) memory. The proposed algorithm consists of selecting a subset of bounding boxes in the image domain, and querying them for the presence of the object of interest by running a pre-trained CNN classifier. The resulting observations are then aggregated and a posterior distribution

show abstract

Section: Counting By Detectionmentioning

confidence: 99%

Cox Processes for Counting by Detection

Rajan

Jedynak

2018

J Math Imaging Vis

View full text Add to dashboard Cite

show abstract

“…Many recent works exploiting point processes have been proposed to address a large variety of computer vision problems [2][3][4][5][6][7][8][9]. The growing interest in these probabilistic models is motivated by the need to manipulate parametric objects interacting in scenes.…”

Section: Point Processes For Vision Problemsmentioning

confidence: 99%

“…Descombes et al [2] propose a point process for counting populations from aerial images, each entity being captured by an ellipse. Ge et al [3] present a point process for a similar application, but dedicated to crowd detection from groundbased photos, for which objects are defined as a set of body shape templates learned from training data. Multi-view images are used by Utasi et al [9] to detect people by a point process in 3D where the objects are specified by cylinders.…”

Section: Point Processes For Vision Problemsmentioning

confidence: 99%

“…These drawbacks explain why industry has been reluctant until now to integrate these mathematical models in their products. Indeed, the point processes presented in Section 1.1 emphasize complex model formulations by proposing objects parametrically sophisticated [3,4], advanced techniques to fit objects to the data [9], and non-trivial spatial interactions between objects [6,8]. However, these works have only slightly addressed the optimization issues from such complex models.…”

Section: Motivationsmentioning

confidence: 99%

“…We propose an efficient mechanism allowing the modifications, creations or removals of objects by taking into account information on the observed scenes. Contrary to the data-driven solutions proposed by [3] and [14], our non-uniform distribution is not built directly from image likelihood, but is created via space-partitioning trees to ensure the sampling parallelization. Efficient GPU implementation -We propose an implementation on GPU which significantly reduces computing times with respect to existing algorithms, while increasing stability and improving the quality of the obtained solution.…”

Section: Contributionsmentioning

confidence: 99%

See 2 more Smart Citations

Efficient Monte Carlo Sampler for Detecting Parametric Objects in Large Scenes

Verdie

Lafarge

2012

Computer Vision – ECCV 2012

View full text Add to dashboard Cite

Point processes have demonstrated efficiency and competitiveness when addressing object recognition problems in vision. However, simulating these mathematical models is a difficult task, especially on large scenes. Existing samplers suffer from average performances in terms of computation time and stability. We propose a new sampling procedure based on a Monte Carlo formalism. Our algorithm exploits Markovian properties of point processes to perform the sampling in parallel. This procedure is embedded into a data-driven mechanism such that the points are non-uniformly distributed in the scene. The performances of the sampler are analyzed through a set of experiments on various object recognition problems from large scenes, and through comparisons to the existing algorithms.

show abstract

Bibliography

2013

Stochastic Geometry for Image Analysis

View full text Add to dashboard Cite

Marked point processes for crowd counting

Cited by 230 publications

References 17 publications

Cox Processes for Counting by Detection

Cox Processes for Counting by Detection

Efficient Monte Carlo Sampler for Detecting Parametric Objects in Large Scenes

Bibliography

Contact Info

Product

Resources

About