Action recognition by dense trajectories

By Heng Wang, Alexander Kläser, Cordelia Schmid, Chenglin Liu. 2011

2,386 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Computer Vision and Pattern Recognition

Citations in the corpus

Two-Stream Convolutional Networks for Action Recognition in Videos (2014, methodology): “Finally, the trajectory feature [29] is computed by stacking the displacement vectors along the trajectory, which corresponds to the trajectory stacking (2).”
Weakly Supervised Action Labeling in Videos under Ordering Constraints (2014, background): “Significant progress towards action recognition in realistic video settings has been achieved in the past few years [22,24,26,30,35].”
Beyond short snippets: Deep networks for video classification (2015, background): “These approaches are able to aggregate local appearance and motion information using hand-crafted features such as Histogram of Oriented Gradients (HOG), Histogram of Optical Flow (HOF), Motion Boundary Histogram (MBH) around spatio-temporal interest points [17], in a dense grid [24] or around dense”
Delving Deeper into Convolutional Networks for Learning Video Representations (2015, methodology): “While previous work has traditionally relied on hand-crafted and task-speciﬁc representations (Wang et al., 2011; Sadanand & Corso, 2012), there is a growing interest in designing general video representations that could help solve tasks in video understanding such as human action recognition,…”
Histogram of Oriented Principal Components for Cross-View Action Recognition (2014, background): “Motion trajectory based action recognition methods [20], [50], [51], [52] are also not reliable in depth sequences [9].”
HOPC: Histogram of Oriented Principal Components of 3D Pointclouds for Action Recognition (2014, neutral): “Motion trajectory based action recognition methods[30,34] are also not reliable in depth sequences [19].”
Skeleton-Based Action Recognition Using Spatio-Temporal LSTM Network with Trust Gates (2017, methodology): “[67], or ConvNet-based features [68], [69]) from the RGB frames, around each body joint as the complementary information.”
PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding (2017, methodology): “Densely tracking points in the optical flow field with more features like Histogram of Oriented Gradient (HOG), Histogram of Flow (HOF) and Motion Boundary Histograms (MBH) encoded by Fisher Vector [24, 42] achieved a good performance.”
Hierarchical Recurrent Neural Encoder for Video Representation with Application to Captioning (2015, background): “Dense Trajectories [39] and its improved version: improved Dense Trajectories [40] have dominated the ﬁled of action recognition and general video classiﬁcation tasks such as complex event detection.”
Video Paragraph Captioning Using Hierarchical Recurrent Neural Networks (2015, methodology): “Alternatively we compute the Dense Trajectories [50] for each video interval and encode them with the Fisher vector [17].”
C3D: Generic Features for Video Analysis (2014, background): “Recently, Wang et al. proposed dense trajectories [50, 51] and later proposed an improved version [52] which is currently the state-of-the-art hand-crafted feature.”
Beyond Gaussian Pyramid: Multi-skip Feature Stacking for action recognition (2014, methodology): “In fact, most of the qualitative improvements to visual analysis can be attributed to the introduction of improved representations, from SIFT [15] to Deep Convolutional Neural Networks [29], STIP [12] to Dense Trajectory [38] .”
Actions ~ Transformations (2015, background): “More recently, trajectory based approaches [51, 31, 29, 50, 18, 32, 25] have shown signiﬁcant improvement in action recognition.”
Finding action tubes (2014, background): “[40] CNN (1/3 spatial, 2/3 motion) Action Tubes”
Jointly Modeling Deep Video and Compositional Text to Bridge Vision and Language in a Unified Framework (2015, methodology): “We apply the state-of-the-art motion descriptor, i.e. dense trajectory (Wang et al. 2011) and object descriptor , i.e. ObjectBank (Li et al. 2011) as visual features.”
Every Moment Counts: Dense Detailed Labeling of Actions in Complex Videos (2015, background): “[34], achieve excellent results on benchmark action recognition datasets.”
Moments in Time Dataset: One Million Videos for Event Understanding (2018, background): “[43] developed dense action trajectories by separating foreground motion from camera motion.”