Detecting Pedestrians Using Patterns of Motion and Appearance

By Paul A. Viola, Michael J. Jones, Daniel Snow. Proceedings Ninth IEEE International Conference on Computer Vision, 2003

2,691 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/669b01f848f7abfbb60f7b1e32d55206942ab090

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

An Integrated Model of Top-Down and Bottom-Up Attention for Optimizing Detection Speed (2006, methodology): “While this approach has been successfully applied to detecting rigid objects such as faces and cars [20, 22, 29] and even pedestrians [17, 30], it is slow and computationally expensive as each classifier (corresponding to every object) is run independently at every window within the image.”
Discriminative Autoencoders for Small Targets Detection (2014, methodology): “Most of the techniques are based on the sliding windows framework, combining descriptors, such as Histogram of Oriented Gradient [3] or Haar Wavelets [11], with powerful discriminative classifier, such as boosting [8] or SVM [9].”
Video-Based Human Behavior Understanding: A Survey (2013, methodology): “The Viola and Jones algorithm, used initially in face detection, can be trained for pedestrian detection [35].”
Spatiotemporal Group Context for Pedestrian Counting (2014, background): “We review state of the arts from two aspects: 1) pedestrian detection and 2) pedestrian counting.”
Pedestrian detection and tracking with night vision (2005, methodology): “838222 by shape and motion with feature vectors obtained by a bank of rectangular filters [36]; stereo-based disparity segmentation and neural network-based recognition [38]; stereo-based object detection and motion analysis [34]; analysis on the vertical pro-”
Ten Years of Pedestrian Detection, What Have We Learned? (2014, methodology): “Brief chronology In 2003, Viola and Jones applied their VJ detector [44] to the task of pedestrian detection.”
Multi-cue onboard pedestrian detection (2009, background): “While psychologists and neuroscientists argue that motion is an important cue for human perception [17] only few computer vision object detectors (e.g. [30, 6]) exploit this fact.”
Generic object recognition with boosting (2006, methodology): “Complementing our approach, Viola et al. [39] use motion information as a second source of information, whereas we use various techniques to describe image intensity information.”
Semantic Amodal Segmentation (2015, methodology): “or existing algorithms. We are currently annotating ˘5000 images from the MS COCO dataset. 1. Introduction In recent years, visual recognition tasks such as image classiﬁcation [17], object detection [30,10,24,13], edge detection [2,7,33], and semantic segmentation [25,9,22] have witnessed dramatic ”
Single-Pedestrian Detection Aided by Multi-pedestrian Detection (2013, methodology): “3) Features under investigation include Haar-like features [32], edgelets [37], shapelets [27], histogram of gradients (HOG) [4], bag-of-words [18], integral histograms [26], color histograms [33], covariance descriptors [31], co-occurrence features [28], local binary patterns [34], color-self-simil”
Robust Multi-resolution Pedestrian Detection in Traffic Scenes (2013, methodology): “Traditional pedestrian detectors usually follow the scale invariant assumption: a scale invariant feature based detector trained at a fixed resolution could be generalized to all resolutions, by resizing the detector [40, 4], image [6, 19] or both of them [11].”
A Two-Stage Template Approach to Person Detection in Thermal Imagery (2005, methodology): “Some examples include the direct use of wavelet features with support vector machines [6], coarse-to-ﬁne edge template matching [3], motion/intensity AdaBoosted classiﬁers [7], and the size/shape of image differencing regions [4].”
Moving Object Detection by Detecting Contiguous Outliers in the Low-Rank Representation (2011, methodology): “Generally, the classifier is built by offline learning on separate datasets [3], [4] or by online learning initialized with a manually labeled frame at the start of a video [5], [6].”
Fast Human Detection Using a Cascade of Histograms of Oriented Gradients (2006, methodology): “Viola et al. [ 14 ] extended their Haar-like wavelets to handle space-time information for moving-human detection.”
Pedestrian Detection: An Evaluation of the State of the Art (2012, background): “Given a static camera, Viola et al. [53] proposed computing Haar-like features on difference images, resulting in large performance gains.”
Human Detection Using Oriented Histograms of Flow and Appearance (2006, methodology): “[23] based on simple spatiotemporal differencing rather than flow.”
DeepID-Net: multi-stage and deformable deep convolutional neural networks for object detection (2014, background): “e of object detection are important. First, features should capture the most discriminative information of object classes. Well-known features include hand-crafted features such as Haar-like features [55], SIFT [32], HOG [9], and learned deep CNN features [46, 29, 23]. Second, deformation models sho”