View-Invariant Representation and Recognition of Actions

By Cen Rao, Alper Yilmaz, Mubarak Shah. 2002

525 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Human-Computer Interaction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

Histogram of Oriented Principal Components for Cross-View Action Recognition (2014, background): “Since HOPC and STK-Distribution capture complimentary information, their combination significantly improves classification accuracy which is over 22% higher than the nearest competitor on both datasets.”
HOPC: Histogram of Oriented Principal Components of 3D Pointclouds for Action Recognition (2014, methodology): “In RGB videos, in order to recognize actions across viewpoint changes, mostly view independent representations are proposed such as view invariant spatio-temporal features [2,20,22,23,25,33].”
Actions sketch: a novel action representation (2005, background): “Single point tracking generates a motion trajectory, and there are several approaches employing motion trajectories for action recognition [8].”
Action recognition from motion capture data using Meta-Cognitive RBF Network classifier (2014, background): “[7] [8] [9] [10] and those based on 3D motion streams on skeleton data, e.”
Coupled Action Recognition and Pose Estimation from Multiple Views (2012, methodology): “matching (Gavrila and Davis 1995; Yacoob and Black 1999; Rao et al. 2002) or with state-space models such as HMMs (Campbell and Bobick 1995).”
Trajectons: Action recognition through the motion analysis of tracked features (2009, methodology): “…trajectons, each trajectory snippet is simply a concatenated vector of (dx, dy) derivatives for the trajectory, whereas in AffineAugmented (AA) trajectons, this vector of derivatives is concatenated with a vector of local affine transforms to describe the motion around each trajectory point.”
Kernel-Based Object Tracking (2003, methodology): “[61] employed kernel tracking in their system for action recognition, while Caenen et al.”
Observing Human-Object Interactions: Using Spatial and Functional Compatibility for Recognition (2009, background): “They discovered that the weight of an object is most robustly estimated, while size and shape are harder to estimate.”
Action recognition in videos acquired by a moving camera using motion decomposition of Lagrangian particle trajectories (2011, background): “In addition, several studies assume that the motion trajectories are already available [5], or they rely on manual annotations [4], or the so-called semi-automatic manner [13].”
Multiple/Single-View Human Action Recognition via Part-Induced Multitask Structural Learning (2015, unknown): “[29] showed that the maxima in spacetime curvature of a 3-D trajectory were preserved in 2-D image trajectories, and therefore the 2-D trajectories can be utilized to capture the view-independent representation of human actions.”
Learning to Recognize Activities from the Wrong View Point (2008, methodology): “Alternatively, one might use an implicit representation of 3D shapes as a set of silhouettes from multiple views [2,26], spatio-temporal volumes derived from such a silhouettes [42,47], or spatio-temporal curvature of 2D trajectory [32].”
View Invariance for Human Action Recognition (2005, background): “Rao et al. (2002) consider the problem of learning and recognizing actions performed by a human hand.”
Action recognition by learning mid-level motion features (2008, background): “[21] track hands, and describe actions based on inflection points of these 2D tracks.”
3D Action Recognition from Novel Viewpoints (2016, background): “…recognition is challenging because significant intra-action variations exist due to changes in viewpoint, illumination, visual appearance (such as color and texture of clothing), scale (due to different human body sizes or distances from the camera), background and speed of performing an action.”
Does Human Action Recognition Benefit from Pose Estimation? (2011, methodology): “The sequences themselves were then classified through exemplar matching [12, 23, 32] or with state-space models such as HMMs [5].”
Summarizing surveillance videos with local-patch-learning-based abnormality detection, blob sequence optimization, and type-based synopsis (2015, methodology): “transmission-based method (NTB) [5]; (2) The kernel density estimation method which directly models probability densities for different activities based on training trajectories (KDE) [5,13]; (3) The dynamic time warping method which detect abnormalities by finding their closest time-warped trajecto”