Temporal Activity Detection in Untrimmed Videos with Recurrent Neural Networks
来源:互联网 发布:vc ado access数据库 编辑:程序博客网 时间:2024/06/05 15:16
团队介绍
作者:
Alberto Montes, Amaia Salvador, Santiago Pascual, Xavier Giro-i-Nieto
作者都来自Universitat Politècnica de Catalunya (UPC)的一所西班牙大学,理工科挺强的,发表在NIPS workshop的一篇文章. 在ActivityNet Challenge 2016取得了不错的
动机
利用C3D[1]能捕捉短时间的空时特征,然后LSTM处理长时间的信息,Untrimmed Videos对进行分类和定位
框架
C3D首先在sports-1M上进行预训练,之后对预处理好的每16帧视频(相邻的视频片段没有交叉)离线提取时空特征,将这些固定的特征作为LSTM的输入,进行每一个片段的分类,每一类对应于一个动作类别(增加background作为一类)。作者在文中也探讨了不同深度,不同宽度的LSTM网络。发现1x512的最浅最窄的最好。The simplest is the best.
tricks
- 对LSTM输出的概率进行均值滤波,使其更平滑,消除异常概率值
- 为了应对背景数据较多的情况,在计算loss时候,给其相对较小的权重
思考改进
如果C3D和LSTM一起训练,微调C3D,重新训练LSTM,效果应该会好一点,但用于参数众多,也可能导致参数过多,导致过拟合
引用
- D. Tran, L. Bourdev, R. Fergus, L. Torresani, and M. Paluri, Learning Spatiotemporal Features with 3D Convolutional Networks, ICCV 2015
阅读全文
0 0
- Temporal Activity Detection in Untrimmed Videos with Recurrent Neural Networks
- T-CNN: Tubelets with Convolutional Neural Networks for Object Detection from Videos
- Supervised Sequence Labelling with Recurrent Neural Networks
- Generating News Headlines with Recurrent Neural Networks
- Chinese Poetry Generation with Recurrent Neural Networks
- Recurrent Neural Networks with Word Embeddings¶
- Chinese Poetry Generation with Recurrent Neural Networks
- Exploring Sparsity in Recurrent Neural Networks
- Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- Text Generation With LSTM Recurrent Neural Networks in Python with Keras
- 论文《Inside-Outside Net: Detecting Objects in Context with skip pooling and Recurrent Neural Networks》
- Gated Recurrent Neural Networks
- Recurrent Neural Networks Tutorial
- Recurrent Neural Networks - collections
- Recurrent Neural Networks regularization
- Recurrent Neural Networks
- Recurrent Neural Networks
- SSN:Temporal Action Detection with Structured Segment Networks
- 进程控制编程
- Android四大组件 BroadCasrReciver
- 如何实现session共享
- bzoj3237[Ahoi2013]连通图 cdq分治+并查集
- C++虚函数和多态继承
- Temporal Activity Detection in Untrimmed Videos with Recurrent Neural Networks
- 补码(为什么按位取反再加一):告诉你一个其实很简单的问题
- Android-Binder系统APP实现
- 关于Prime算法的从入门到升天的讲解(带模板)
- Java的变量和数据类型
- 【bzoj3669】[Noi2014]魔法森林
- sklearn因子分析(python)
- poj3349——Snowflake Snow Snowflakes
- 解决Chrome下更新到最新的Adobe Flash Player仍然出现点击无法播放Flash的情况