Discriminative virtual views for cross-view action recognition

By Ruonan Li, Todd Zickler. 2012

191 citations recorded by the source corpus.

We propose an approach for cross-view action recognition by way of `virtual views' that connect the action descriptors extracted from one (source) view to those extracted from another (target) view. Each virtual view is associated with a linear transformation of the action descriptor, and the sequence of transformations arising from the sequence of virtual views aims at bridging the source and target views while preserving discrimination among action categories. Our approach is capable of operating without access to labeled action samples in the target view and without access to corresponding action instances in the two views, and it also naturally incorporate and exploit corresponding instances or partial labeling in the target view when they are available. The proposed approach achieves improved or competitive performance relative to existing methods when instance correspondences or target labels are available, and it goes beyond the capabilities of these methods by providing some level of discrimination even when neither correspondences nor target labels exist.

Subfields: Biomedical Engineering, 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 Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Histogram of Oriented Principal Components for Cross-View Action Recognition (2014, background): “Keywords: Spatio-temporal keypoint, pointcloud, view invariance.”
HOPC: Histogram of Oriented Principal Components of 3D Pointclouds for Action Recognition (2014, background): “Another approach is to find a view independent latent space [7,8,12,14] in which features extracted from the actions captured at different view points are directly comparable.”
Cross-View Action Modeling, Learning, and Recognition (2014, methodology): “In contrast, although the state-of-the-art local-feature-based cross-view action recognition methods [10, 11] are relatively robust to viewpoint changes, the overall accuracy of these methods is not very high, because the local features are not enough to discriminate the subtle differences of the ac”
3D Pose from Motion for Cross-View Action Recognition via Non-linear Circulant Temporal Encoding (2014, unknown): “Li and Zickler [14] and more recently Zhang et al.”
Learning View-Invariant Sparse Representations for Cross-View Action Recognition (2013, methodology): “For fair comparison to [6, 20, 15], we use three evaluation modes: (1) unsupervised correspondence mode; (2) supervised correspondence mode ; (3) partially labeled mode.”
Learning a Deep Model for Human Action Recognition from Novel Viewpoints (2016, methodology): “Besides the limitations of employing linear transformation functions between views, existing cross-view action recognition methods [15], [18], [20], [21], [22], [23] are either not”
View Adaptive Neural Networks for High Performance Skeleton-Based Human Action Recognition (2018, background): “Some attempts have been made in previous works to overcome the view variation problem for robust action recognition [3], [8], [9], [14], [15], [19], [24], [26], [31], [36], [37], [39], [40], [50], [52], [53], [60].”
Cross-View Action Recognition via a Continuous Virtual Path (2013, methodology): “Li et al ’s work [15], which also explored the idea of using virtual view to overcome the problem of view changes, is close to ours.”
Movement Pattern Histogram for Action Recognition and Retrieval (2014, methodology): “For the action retrieval task, we compare retrieval results against several BOF representations: cuboid [5], self-similarity matrix (SSM) [10], shapeflow [25], dense trajectories [26], and cuboid+shapeflow (used in [17,16,30]).”
A Review of Human Activity Recognition Methods (2015, methodology): “Space-time methods involve activity recognition methods, which represent human activities as a set of spatiotemporal features (Shabani et al., 2011; Li and Zickler, 2012) or trajectories (Li et al., 2012; Vrigkas et al., 2013).”
Deep Domain Adaptation in Action Space (2018, background): “There are few studies [6, 19, 31] on cross-view action recognition and a few on heterogeneous domain adaptation [8, 9].”
Cross-View Action Recognition over Heterogeneous Feature Spaces (2013, methodology): “Different from [8], our method can handle the cross-view action recognition when the actions are represented by heterogeneous features in source and target views.”