================================================================================
Convolutional Two-Stream Network Fusion for Video Action Recognition
This repository contains the code for our CVPR 2016 paper:
Christoph Feichtenhofer, Axel Pinz, Andrew Zisserman
"Convolutional Two-Stream Network Fusion for Video Action Recognition"
in Proc. CVPR 2016
If you find the code useful for your research, please cite our paper:
@inproceedings{feichtenhofer2016convolutional,
title={Convolutional Two-Stream Network Fusion for Video Action Recognition},
author={Feichtenhofer, Christoph and Pinz, Axel and Zisserman, Andrew},
booktitle={Conference on Computer Vision and Pattern Recognition (CVPR)},
year={2016}
}
Requirements
The code was tested on Ubuntu 14.04 and Windows 10 using MATLAB R2015b and NVIDIA Titan X or Z GPUs.
If you have questions regarding the implementation please contact:
Christoph Feichtenhofer
================================================================================
Setup
-
Download the code
git clone --recursive https://github.com/feichtenhofer/twostreamfusion -
Compile the code by running
compile.m.- This will also compile a modified (and older) version of the MatConvNet toolbox. In case of any issues, please follow the installation instructions on the MatConvNet homepage.
-
Edit the file cnn_setup_environment.m to adjust the models and data paths.
-
Download pretrained model files and the datasets, linked below and unpack them into your models/data directory.
- Optionally you can pretrain your own twostream models by running
cnn_ucf101_spatial();to train the appearance network stream.cnn_ucf101_temporal();to train the optical flow network stream.
- Run
cnn_ucf101_fusion();this will use the downloaded models and demonstrate training of our final architecture on UCF101/HMDB51.- In case you would like to train on the CPU, clear the variable
opts.train.gpus - In case you encounter memory issues on your GPU, consider decreasing the
cudnnWorkspaceLimit(512MB is default)
- In case you would like to train on the CPU, clear the variable
Pretrained models
- Download our baseline networks trained on UCF101 here:
Data
Pre-computed optical flow images and resized rgb frames for the UCF101 and HMDB51 datasets
Use it on your own dataset
- Our Optical flow extraction tool provides OpenCV wrappers for optical flow extraction on a GPU.
87 Dec 16, 2022
114 Nov 27, 2022
2 Jan 29, 2022
15 Nov 30, 2022
171 Dec 26, 2022
20 Jan 3, 2023
1.1k Dec 27, 2022
68 Jan 3, 2023
1.1k Dec 25, 2022
80 Dec 8, 2022
479 Dec 26, 2022
267 Dec 17, 2022
8 Apr 16, 2022
70 Nov 4, 2022
67 Jan 3, 2023
27 Jul 20, 2022
368 Jan 6, 2023
123 Dec 13, 2022