Action Recognition from Arbitrary Views using 3D Exemplars

By Daniel Weinland, Edmond Boyer, Rémi Ronfard. 2007

519 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild (2012, background): “…action recognition datasets suffer from two disadvantages: 1) The number of their classes is typically very low compared to the richness of performed actions by humans in reality, e.g. KTH [11], Weizmann [3], UCF Sports [10], IXMAS [12] datasets includes only 6, 9, 9, 11 classes respectively.”
HMDB: A large video database for human motion recognition (2011, background): “State-of-the-art performance on these datasets is now near ceiling and thus there is a need for the design and creation of new benchmarks.”
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): “Some methods infer the 3D scene structure and use geometric transformations to achieve view invariance [5, 9, 15, 26, 39].”
Unstructured human activity detection from RGBD images (2011, methodology): “We have also made the dataset and code available open-source at: http://pr.cs.cornell.edu/humanactivities”
Discriminative virtual views for cross-view action recognition (2012, background): “…three-dimensional scene structure so that the derived action descriptors can be adapted from one view to another through geometric reasoning [29, 26, 15, 10, 4], while another is to search for spatio-temporal features of a video sequence that are insensitive to changes in view angle [17, 21,…”
Cross-View Action Recognition via Transferable Dictionary Learning (2016, methodology): “For example, a binary motion energy image was introduced in [6] and [43] by aggregating differences of silhouettes between subsequent frames of an action video.”
Learning View-Invariant Sparse Representations for Cross-View Action Recognition (2013, methodology): “We evaluated our approach for both cross-view and multi-view action recognition on the IXMAS multi-view dataset [28].”
MuHAVi: A Multicamera Human Action Video Dataset for the Evaluation of Action Recognition Methods (2010, methodology): “Later, they [12] proposed an exemplar based HMM method that also works with one test camera, where human pose estimations are combined probabilistically to recognize the action.”
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].”
View-Invariant Action Recognition Using Latent Kernelized Structural SVM (2012, methodology): “Some existing approaches address the view-invariant action recognition by using epipolar geometry [6] [7] or a full 3D reconstruction [8] [9].”
Cross-view action recognition via view knowledge transfer (2011, methodology): “We test our approach on the IXMAS multi-view action data set [ 6 ] which contains eleven daily-live actions.”
Recognizing realistic actions from videos “in the wild” (2009, unknown): “Local Motion Features: Since the bag of video-words (BOV) approach does not require background subtraction and object tracking [ 3 , 5], and can cope with certain camera motion and illumination changes, it is receiving increasing attention in generic action recognition [8,9,10, 11,12,29].”
Deep multi-view learning methods: A review (2021, background): “WVU [204] Video 4 10 650 2011 Ten kinds of human actions captured by 8 embedded camera networks; Complete overlapping coverage is provided for 8 different perspectives/views; IXMAS[205] Video 5 10 1,320 2007 It is a video dataset containing 11 human actions, the actions of each category is conducted”
Reshaping Visual Datasets for Domain Adaptation (2013, background): “Instead, standard image/video collection is a product of trying to ensure coverage of the target category labels on one hand, and managing resource availability on the other.”
Dense Trajectories and Motion Boundary Descriptors for Action Recognition (2013, methodology): “IXMAS 10 (Weinland et al, 2007) is a dataset recorded by cameras from ﬁve different viewpoints, as shown in the ﬁfth row of Figure 8.”
A survey of video datasets for human action and activity recognition (2013, background): “Finally, Section 5 presents the conclusion of this survey.”