Temporal Convolutional Networks 时序卷积网络 学习笔记

1.  Human Action Recognition using Factorized Spatio-Temporal Convolutional Networks    ICCV2015 

把时序数据的三维卷积分解成先2维空间卷积 (spatial convolutional layer), 再1维的时间卷积(temporal convolutional layers)。

视频片段切割:剪裁数据增强,固定时间段切割。

Vdiff:时间相隔dt的画面之间的diff,以捕捉短程信息。与V的长程信息一起来预测。

ImageNet预训练:2-D空间卷积可以允许使用图片来进行预训练。

SCI based score fusion:多个结果不平均对待,给可信度最高的结果更高的权重来得到更好的结果。

可以参考博客:https://blog.csdn.net/zzmshuai/article/details/84880257

2. Temporal Convolutional Networks for Action Segmentation and Detection   CVPR2017

EncoderDecoder TCN

Dilated TCN

结合了causal convolution, residual connection 和 dilation convolution.

参考博客:https://juejin.im/entry/5b04dac96fb9a07ab979d665/

3.  An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling    

参考博客:http://nooverfit.com/wp/%E6%97%B6%E9%97%B4%E5%8D%B7%E7%A7%AF%E7%BD%91%E7%BB%9Ctcn-%E6%80%BB%E7%BB%93%EF%BC%9A%E6%97%B6%E5%BA%8F%E6%A8%A1%E5%9E%8B%E4%B8%8D%E5%86%8D%E6%98%AF%E9%80%92%E5%BD%92%E7%BD%91%E7%BB%9Crnn/?yyue=a21bo.50862.201879

4. Spatio-Temporal Graph Convolutional Networks: A Deep Learning Framework for Traffic Forecasting   IJCAI2018

参考博客:https://davidham3.github.io/blog/2018/05/10/spatio-temporal-graph-convolutional-networks-a-deep-learning-framework-for-traffic/

 

代码(Source Code):

论文3作者提供的pytorch版本:https://github.com/locuslab/TCN

keras版本:https://github.com/philipperemy/keras-tcn

Tensorflow版本:https://github.com/Baichenjia/Tensorflow-TCN

你可能感兴趣的:(深度学习)