Extract Video Feature (PyTorch)

Pytorch implementation of extracting frame-level features of video by a 2D CNN(ResNet-18). This project is made by Shengeng Tang. Hefei University of Technology, Ph.D candidate.

In this project, we (1) first split the video into frames, (2) then extract the frame-level features, and (3) finally combine the frame-level features into clip-level features for alignment and fusion with other features.

Requirements

See the installation instruction for a step-by-step installation guide. See the server instruction for server settup.

• Install cuda-8.0
• Install cudnn v5.1
• Download PyTorch for python-2.7 and clone the repository.

pip install http://download.pytorch.org/whl/cu80/torch-0.1.12.post2-cp27-none-linux_x86_64.whl
pip install torchvision
git clone https://github.com/tangshengeng/extract-video-feature_PyTorch.git

Split the Video into Frames

python video2frames.py

Extract the Features

python extract_features.py

Combine the Features

python framefeats2clipfeats.py