Paper address: https://openaccess.thecvf.com/content_cvpr_2018/papers/Sun_Optical_Flow_Guided_CVPR_2018_paper.pdf
This repo contains the implementation code for the paper:
Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition, Shuyang Sun, Zhanghui Kuang, Lu Sheng, Wanli Ouyang, Wei Zhang, CVPR 2018.
prerequisites
OpenCV 2.4.12
OpenMPI 1.8.5 (installed with multithreading enabled)
CUDA 7.5
CUDNN 5
data preparation
The data of UCF-101 and HMDB-51 can be prepared by referring to the documentation of the TSN project.
How to build
For training, first modify the file make_train.sh, fill in your own lib path, and run sh make_train.sh, the script will automatically build caffe for you.
For testing, you can simply run make pycaffe to have everything ready.
train
Before starting training