README.md

Vieo Action Recognition

This is our project for building the Video Action Recognition.

Dataset deployment

Dataset deployment steps：

• Make video-action-recognition/data directory.

• Download HMDB51 to the data directory– About 2GB for a total of 7,000 clips distributed in 51 action classes. Add use unrar x xxx.rar to extract all .rar file in this dataset. Finally, we have an video-action-recognition/data with directory tree structure like this:

  data
  └── HMDB51
      ├── split
      │   ├── README
      │   ├── testTrainMulti_7030_splits
      │   └── test_train_splits.rar
      └── video
          ├── brush_hair
          ├── cartwheel
          ├── catch
          ...

• Run dataset/dataset_list_maker.py to create annotation list file.

  python dataset/dataset_list_maker.py data/HMDB51/
  

• At last, video-action-recognition/data directory tree structure will be like this:

  data
  └── HMDB51
      ├── meta.txt
      ├── split
      │   ├── README
      │   ├── testTrainMulti_7030_splits
      │   └── test_train_splits.rar
      ├── test_list.txt
      ├── train_list.txt
      └── video
          ├── brush_hair
          ├── cartwheel
          ├── catch
          ...
  

Model training

Resnet_a models:

• Go to the current directory:cd xxx/video-action-reconigtion
• Run tensorboard: bash tensorboard/tensorboard.sh [port]. e.g bash tensorboard/tensorboard.sh 7788
• Start training with bash experiments/scripts/train_resnet_a.sh