Human Activity Recognition with Metric Learning

By Du Tran, A. Sorokin. European Conference on Computer Vision, 2008

370 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/a73c9533743da02a649e4d22fbe3bcde6f834cc8

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

Learning View-Invariant Sparse Representations for Cross-View Action Recognition (2013, methodology): “For the global feature, we extract shape-flow descriptors introduced in [26] and learn a codebook of size 500 by k-means clustering on these shape-flow descriptors.”
Real-time Action Recognition by Spatiotemporal Semantic and Structural Forests (2010, methodology): “Since these methods [26, 28, 30] encode holistic structures with respect to a reference position e.”
Making Action Recognition Robust to Occlusions and Viewpoint Changes (2010, background): “Most state-of-the-art approaches compute image-sequence descriptors based on variants of either sparse interest points [3,11,17,20,24] or dense holistic features [9,13,19,22,23].”
Transfer Learning in a Transductive Setting (2013, background): “Other directions to improve the quality of the intermediate representation include integrating metric learning [31, 16] or online methods [10] which we defer to future work.”
Spatiotemporal Deformable Part Models for Action Detection (2013, background): “, [5, 22, 26]), these all perform recognition through silhouettes.”
Human Focused Video Description (2011, background): “One by Tran and Sorokin contained transitive actions such as standing to sitting and jumping from sitting [21].”
Object, Scene and Actions: Combining Multiple Features for Human Action Recognition (2010, neutral): “Tran, et al. [30] work on YouTube Badminton videos.”
Web-Based Classifiers for Human Action Recognition (2012, methodology): “Amongst the current literature, some approaches require static or uniform backgrounds and easily extractable silhouettes [2], [7] and some approaches work with more realistic videos in the presence of background clutter [37]–[41].”
Human detection from images and videos: A survey (2016, background): “A straightforward potential task is human segmentation which is important for human shape registration [34], human gait analysis [164, 18] and action recognition [13, 153, 60] using human silhouette.”
Action recognition from motion capture data using Meta-Cognitive RBF Network classifier (2014, neutral): “Tran and Sorokin [19] used motion context and a learned nearest neighbor metric for classifying actions in YouTube videos of sporting events.”
Metric Learning: A Survey (2013, result): “A recent paper [72] considered the application of LMNN to the problem of human activity recognition in video sequences, and showed superior results with the learned metrics as compared to a number of other baselines.”
Evaluation of video activity localizations integrating quality and quantity measurements (2014, background): “Consequently, we see an early drop in performance varying any one of the thresholds, t sr = 1 or t sp = 1 .”
The Action Similarity Labeling Challenge (2012, neutral): “UIUC1 [11] 14 532 Laboratory: 8 actors, single view, extensive repetitions Resolution 400px”
Deep transfer metric learning (2015, methodology): “successfully applied in visual analysis applications such as face recognition [14, 16], image classification [9, 32], human activity recognition [31], person re-identification [19] and visual search [26].”
Discriminative virtual views for cross-view action recognition (2012, methodology): “There are many possible choices for the classifier, and in the experiments we use the Multiple Kernel Learning SVM (MKL-SVM) [20].”
Modeling Scene and Object Contexts for Human Action Retrieval With Few Examples (2011, background): “Torralba et al. [15] introduced a framework for context modeling based on the correlation between overall statistics of low-level image features and objects in the images.”
A Vision-Based System for Intelligent Monitoring: Human Behaviour Analysis and Privacy by Context (2014, background): “In the case of [23], both MHI and the frame descriptor from Tran and Sorokin [24] are tested.”
Joint segmentation and classification of human actions in video (2011, background): “Tran & Sorokin [23] used motion context and a learned nearest neighbor metric for classifying actions in YouTube videos of sporting events.”
Hierarchical Clustering Multi-Task Learning for Joint Human Action Grouping and Recognition (2017, neutral): “1 Tran [38] 80.”