Actions sketch: a novel action representation

By Alper Yilmaz, Mubarak Shah. 2005

506 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

HOPC: Histogram of Oriented Principal Components of 3D Pointclouds for Action Recognition (2014, methodology): “Some methods infer the 3D scene structure and use geometric transformations to achieve view invariance [5,9,15,26,39].”
Scene Flow to Action Map: A New Representation for RGB-D Based Action Recognition with Convolutional Neural Networks (2017, methodology): “magetotakeadvantageoftheavailablepre-trainedmodelsfor standardConvNetsarchitecturewithouttrainingmillionsof parameters afresh. There are several ways to encode the video sequences into dynamic images [4, 5, 80, 38, 78, 67, 68,3],buthowtoencodethesceneﬂowvectorsintoonedynamicimagestillneedstobeexplor”
Going deeper into action recognition: A survey (2016, background): “Yilmaz and Shah propose to identify actions based on the differential properties of the Space-Time Volume (STV) Yilmaz and Shah (2005).”
Learning a Deep Model for Human Action Recognition from Novel Viewpoints (2016, methodology): “To overcome this problem, some techniques infer 3D scene structure and use geometric transformations to achieve view invariance [3], [24], [25], [26], [27], [28], [29], [30], [31], [32].”
Latent Multitask Learning for View-Invariant Action Recognition (2013, background): “One way to achieve view invariance could be to reason about a 3D layout of the scene, or 3D volume of the human body, so that the video features can be adapted from one view to another through geometric transformations [29, 28, 23, 18, 14].”
Chaotic Invariants for Human Action Recognition (2007, background): “Some leading representations are: learned geometrical models of human body parts [19, 20], space-time pattern templates [12, 16]), 3D information ([17]), shape or form features [21, 13]), interest point based representation [23], motion/optical flow patterns [24, 18] and volumetric features [22, 15]”
Learning View-Invariant Sparse Representations for Cross-View Action Recognition (2013, background): “In order to accurately recognize human actions, most existing approaches focus on developing different discriminative features, such as spatio-temporal interest point (STIP) based features [32, 2, 13, 19], shape [16, 3, 21] and optical flow based features [5, 17, 16].”
A survey on activity recognition and behavior understanding in video surveillance (2013, methodology): “Rao and Shah [147] proposed a viewinvariant trajectory matching method to learn human actions without any training model and to avoid ambiguity of action and their trajectories.”
Human Action Recognition and Prediction: A Survey (2018, methodology): “Another method to describe shape and motion was present d in [87].”
Hierarchical spatio-temporal context modeling for action recognition (2009, background): “Therefore the action recognition algorithms [22, 23], which aimed to achieve the robustness to viewpoint change based on geometric reconstruction, are doomed due to their reliance on exact knowledge of the object contours or multiple view geometries, which are prone to errors in unconstrained videos”
Recognizing human actions by attributes (2011, background): “More recently, great progress has been made by introducing more descriptive action representations such as space-time pattern templates[1, 21], 2D shape matching [33, 7], optical flow patterns [6, 33], trajectory-based representation [27], and bag-of-video-words [24, 19, 20].”
Human Action Recognition From Various Data Modalities: A Review (2020, background): “[71] introduced a spacetime volume that is a sequence of 2D contour.”