【CV-Action Recognition】论文目录

2006----Notes on Convolutional Neural Networks

2011----Large displacement optical flow_ descriptor matching in variational motion estimation

2013----3D Convolutional Neural Networks for Human Action Recognition

2013----iDT----Action Recognition with Improved Trajectories

2014----C3D----Learning Spatiotemporal Features with 3D Convolutional Networks

2014----Large-scale Video Classification with Convolutional Neural Networks

2014----Two-Stream Convolutional Networks for Action Recognition in Videos

2015----A Key Volume Mining Deep Framework for Action Recognition

2015----Deep Multi Scale Video Prediction Beyond Mean Square Error

2015----Dense trajectories and motion boundary descriptors for action recognition

2015----Large-scale Video Classification with Convolutional Neural Networks

2016----A Torch Library for Action Recognition and Detection Using CNNs and LSTMs

2016----Anticipating Visual Representations from Unlabeled Video

2016----Convolutional Two-Stream Network Fusion for Video Action Recognition

2016----Delving Deeper into Convolutional Networks for Learning Video Representations

2016----Flow-Guided Feature Aggregation for Video Object Detection

2016----Long-term temporal convolutions for action recognition

2016----Spatiotemporal Residual Networks for Video Action Recognition

2016----Spot On_Action Localization from Pointly-Supervised Proposals

2016----The THUMOS Challenge on Action Recognition for Videos

2016----UntrimmedNets for Weakly Supervised Action Recognition and Detection

2016----Real-time Action Recognition with Enhanced Motion Vector CNNs

2017----Action Tubelet Detector for Spatio-Temporal Action Localization

2017----Aggregating Frame-level Features for Large-Scale Video classification

2017----Am I Done-Predicting Action Progress in Videos

2017----AMTnet_Action-Micro-Tube regression by end-to-end trainable deep architecture

2017----Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection

2017----Generic Tubelet Proposals for Action Localization

2017----Hidden Two-Stream Convolutional Networks for Action Recognition

2017----Incremental Tube Construction for Human Action Detection

2017----Multi-Task Clustering of Human Actions by Sharing Information

2017----Object Detection in Videos with Tubelet Proposal Networks

2017----Online Real-time Multiple Spatiotemporal Action Localisation and Prediction

2017----Quo Vadis, Action Recognition_A New Model and the Kinetics Dataset

2017----Single Shot Temporal Action Detection

2017----Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos

2017----On the Integration of Optical Flow and Action Recognition

在行为识别领域,比较主流的算法有two-streams,3D convolutions 和RNN,尤其以two-streams算法
性能良好。而在two-streams算法中一般集成光流计算,但是为什么光流算法有效?光流的计算精度和
行为识别的计算精度相关?有比光流更好的行为识别表示?
1、作者认为two-streams 的光流不是表示运动信息,而是表示外观不变性。
2、用行为识别分类误差来训练(fine tune)光流比起用EPE误差来能获得更好的行为识别效果。



返回CV总目录

编辑于 2018-01-05

文章被以下专栏收录