A Comparative Study of Encoding, Pooling and Normalization Methods for Action Recognition

By Xingxing Wang, Limin Wang, Yu Qiao. 2012

145 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

AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks for Human Action Recognition in Videos (2016, methodology): “Later works improved upon this pipeline in several ways [23, 47] by using dense sampling for feature extractions [41], describing trajectories instead of 3D points [13, 39], and using better pooling and encoding methods [47, 25, 23].”
Towards Good Practices for Action Video Encoding (2014, unknown): “• VLAD is maxent feature learning.”
Latent Hierarchical Model of Temporal Structure for Complex Activity Classification (2014, methodology): “Among these statistical models, Bag of Visual Words (BoVW) is a common choice in action recognition [26].”
RNN Fisher Vectors for Action Recognition and Image Annotation (2015, methodology): “The new representation is sensitive to ordering and therefore mitigates the disadvantage of using the standard Fisher Vector representation.”
Multi-view Super Vector for Action Recognition (2014, methodology): “Among them, FV [23] and its variant VLAD [14], initially designed for image classification, have been shown to achieve promising performances in several action recognition datasets [37, 21, 12].”
Action and Event Recognition with Fisher Vectors on a Compact Feature Set (2013, unknown): “While recently FVs have been explored by others for action recognition [39, 44], we are the first to use them in a large, diverse, and comprehensive evaluation.”
Motionlets: Mid-level 3D Parts for Human Motion Recognition (2013, background): “These properties make local descriptors robust to intra-class variability and deformation to a certain degree.”
Video Action Detection with Relational Dynamic-Poselets (2014, methodology): “Most of the research efforts have been devoted to the problem of action recognition using the Bag of Visual Words (BoVW) framework or variants thereof [29,24,11].”
A Review on Human Activity Recognition Using Vision-Based Method (2017, methodology): “Encoding feature is a key step for constructing BoVW representation and utilizing an appropriate encoding method can signi ﬁ cantly improve the recognition accuracy [56].”
Action Recognition with Stacked Fisher Vectors (2014, methodology): “Recently, advanced feature encoding methods have been introduced for action recognition, such as soft-assignment [21,36], vector of locally aggregated descriptors [9,8], and Fisher coding [24,36,32].”
Large Margin Dimensionality Reduction for Action Similarity Labeling (2014, background): “, hard assignment [12]) can steadily improve action recognition performance [15].”