action recognition 实现方式总结

1. 早期尝试
   Large-scale Video Classification with Convolutional Neural Networks(2014)

GitHub：https://cs.stanford.edu/people/karpathy/deepvideo/

2DCNN, 能不能自动的捕捉运动信息?

• single frame

• stacked frames

2. 双流法
   Two-Stream Convolutional Networks for Action Recognition in Videos(2014)
   双流架构:

   • 空域网络(spatial networks) 执行物体识别

   • 时域网络(temporal networks) 识别运动信息

TSN: Temporal Segment Networks: Towards Good Practices for Deep Action Recognition(2016)

3. (光流的作用)
   On the Integration of Optical Flow and Action Recognition(2017) (2018CVPR)

4. C3D
   Learning Spatiotemporal Features with 3D Convolutional Networks(2015)

5. I3D(Two-Stream Inflated 3D ConvNet)
   Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

action recognition论文汇总