Human Detection Using Oriented Histograms of Flow and Appearance

By Navneet Dalal, Bill Triggs, Cordelia Schmid. 2006

1,965 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Delving into Egocentric Actions (2015, positive): “Note our motion features are a subset from IDT with trajectory features, HoF, MBHx and MBHy.”
Two-Stream Convolutional Networks for Action Recognition in Videos (2014, methodology): “The best performance in the trajectory-based pipeline was achieved by the Motion Boundary Histogram (MBH) [8], which is a gradient-based feature, separately computed on the horizontal and vertical components of optical flow.”
Describing Videos by Exploiting Temporal Structure (2015, methodology): “Each cuboid is represented by concatenating the histograms of oriented gradients, oriented flow and motion boundary (HoG, HoF, MbH) [9, 40] with 33 bins.”
SlowFast Networks for Video Recognition (2018, methodology): “These methods, including histograms of flow [33], motion boundary histograms [6], and trajectories [53], had shown competitive performance for action recognition before the prevalence of deep learning.”
C3D: Generic Features for Video Analysis (2014, methodology): “Laptev and Lindeberg [31] proposed spatio-temporal interest points (STIPs) by extending Harris corner detectors to 3 D. His-togram of Oriented Gradients (HOG) [8] and Histogram of Optical Flows (HOF) [9] are normally extracted at these STIPs and used as video features.”
Actions ~ Transformations (2015, background): “Most successful examples are 3D Histogram of Gradient (HOG3D) [22], Histogram of Optical Flow (HOF) [27], and Motion Boundary Histogram [3].”
Finding action tubes (2014, background): “For the task of action classification, recent approaches use features based on shape (e.g. HOG [5], SIFT [28]) and motion (e.g. optical flow, MBH [6]) with high order encodings (e.g. Bag of Words, Fischer vectors) and train classifiers (e.g. SVM, decision forests) to make action predictions.”
P-CNN: Pose-Based CNN Features for Action Recognition (2015, methodology): “For each trajectory, 4 descriptors are computed in the aligned spatio-temporal volume: HOG [9], HOF [22] and MBH [10].”
Learning Human Pose Estimation Features with Convolutional Networks (2013, methodology): “[37] use HOG [12] features and boosting for learning a parameter sensitive hash function.”
Unsupervised Learning using Sequential Verification for Action Recognition (2016, background): “Typical features like STIP [17], HOG3D [18], HOF [19], MBH [20] etc. capture spatiotemporal features in a temporal grid-like structure.”
Asynchronous Temporal Fields for Action Recognition (2016, methodology): “Understanding activities and actions has an extensive history [32, 59, 22, 17, 23, 2, 26, 56, 29, 21].”
Geometric Context from Videos (2013, methodology): “This is similar to the approach of [4], which has been shown to to be helpful to the task of object detection in videos.”
Exploiting Image-trained CNN Architectures for Unconstrained Video Classification (2015, methodology): “For the IDT descriptors, we concatenated histogram of oriented gradients [4], histogram of ﬂow [5], and motion boundary histograms [16] descriptors extracted along the estimated motion trajectory.”
Multi-View Clustering and Semi-Supervised Classification with Adaptive Neighbours (2017, methodology): “For example, in visual data, an image could be represented by different descriptors, such as SIFT (Lowe 2004), HOG (Dalal, Triggs, and Schmid 2006), GIST (Oliva and Torralba 2001), LBP (Ojala, Pietikäinen, and Mäenpää 2002); in ResearchIndex network, the keywords of a specific paper and its cita”
A database for fine grained activity detection of cooking activities (2012, methodology): “We pick a recently suggested approach [38] which extracts histograms of oriented gradients (HOG), flow (HOF) [16], and motion boundary histograms (MBH) [7] around densely sampled points, which are tracked for 15 frames by median filtering in a dense optical flow field.”
AnomalyNet: An Anomaly Detection Network for Video Surveillance (2019, background): “For example, histogram of oriented gradients (HOG) [8], 3D spatiotemporal gradient [9], and the histogram of oriented flows (HOF) [10] have been extensively used in [6] and [11]–[13].”