Trajectory Convolution for Action Recognition
Explore this paper's citation graph
Summary
This work proposes a new CNN architecture TrajectoryNet, which incorporates trajectory convolution, a new operation for integrating features along the temporal dimension, to replace the existing temporal convolution.
- Type
- article
- Published
- 2018-01-01
- Cited by
- 96
- References
- 52
- OpenAlex
- https://openalex.org/W2891446678
- Semantic Scholar
- https://api.semanticscholar.org/CorpusID:54040787
Keywords
Convolution (computer science), Computer science, Leverage (statistics), Action recognition, Trajectory
References
- UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild
- FlowNet: Learning Optical Flow with Convolutional Networks
- Learning Spatiotemporal Features with 3D Convolutional Networks
- Action recognition with trajectory-pooled deep-convolutional descriptors
- Aggregating Local Image Descriptors into Compact Codes
- Evaluation of Local Spatio-temporal Features for Action Recognition
- Large-Scale Video Classification with Convolutional Neural Networks
- On Space-Time Interest Points
- Secrets of optical flow estimation and their principles
- Visual perception of biological motion and a model for its analysis
- Hierarchical spatio-temporal context modeling for action recognition
- ImageNet Large Scale Visual Recognition Challenge
- An Iterative Image Registration Technique with an Application to Stereo Vision
- Action recognition by dense trajectories
- Learning realistic human actions from movies
- Two-Stream Convolutional Networks for Action Recognition in Videos
- Histograms of oriented gradients for human detection
- Activity recognition using the velocity histories of tracked keypoints
- Deep Residual Learning for Image Recognition
- A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation
Cited by
- EDVR: Video Restoration With Enhanced Deformable Convolutional Networks
- Action Recognition With Spatial-Temporal Discriminative Filter Banks
- Mobile Video Action Recognition
- Recurrent Space-time Graph Neural Networks
- Spatio-temporal deformable 3D ConvNets with attention for action recognition
- The Sound of Motions
- SAST: Learning Semantic Action-Aware Spatial-Temporal Features for Efficient Action Recognition
- Learning Efficient Video Representation with Video Shuffle Networks
- Gate-Shift Networks for Video Action Recognition
- V4D: 4D Covolutional Neural Networks for Video-level Representations Learning
- Action Recognition Using Deep 3D CNNs with Sequential Feature Aggregation and Attention
- V4D: 4D Convolutional Neural Networks for Video-level Representation Learning
- An End to End Framework With Adaptive Spatio-Temporal Attention Module for Human Action Recognition
- TEA: Temporal Excitation and Aggregation for Action Recognition
- Temporal Reasoning Graph for Activity Recognition
- Human Interaction Recognition Based on Whole-Individual Detection
- A novel multi-stream method for violent interaction detection using deep learning
- Adaptive Interaction Modeling via Graph Operations Search
- Video trajectory analysis using unsupervised clustering and multi-criteria ranking
- Spatiotemporal Fusion in 3D CNNs: A Probabilistic View