ActionViewer

This repository is the Action Viewer for the ICVL Action Dataset. More information can be find on our paper. Following image indicates a hierarchical action structure. There are 4 layers, posture, locomotiom, gesture, and event layer. So, one action is represented 4 sub-action labels.

Requirements

opencv 3.3.1
numpy 1.15.4
xlsxwriter 0.9.6

Action Viewer Demo

Documentation

Directory Hierarchy

.
│   ActionViewer
│   ├── DB
│   │   ├── AR-002-12-20151017-01-09.mkv
│   │   ├── AR-002-12-20151017-01-09.txt
│   │   ├── AR-003-12-20151024-02-04.avi
│   │   └── AR-003-12-20151024-02-04.txt
│   ├── Icon
│   │   ├── Bending.bmp
│   │   ├── Bicycling.bmp
│   │   ├── Falling.bmp
│   │   ├── Littering.bmp
│   │   ├── Lying.bmp
│   │   ├── Nothing.bmp
│   │   ├── Others.bmp
│   │   ├── Phoning.bmp
│   │   ├── Running.bmp
│   │   ├── Sitting.bmp
│   │   ├── Smoking.bmp
│   │   ├── Standing.bmp
│   │   ├── Stationary.bmp
│   │   ├── Texting.bmp
│   │   └──  Walking.bmp
│   ├── ICVL_action_structure.py
│   ├── ICVL_data_reader.py
│   ├── main.py
│   └──  write2excel.py

Run Action-Viewer

Run main.py in the ActionViewer.

python main.py

--resize_ratio: resize ratio for input frame, default: 1.0
--interval_time: interval time between two frames, default: 20
Note: main.py read all of the videos and the corresponing txt files.

ICVL Dataset

Click here to download ICVL dataset. Please cite our following paper to use ICVL dataset.

Citation

@article{jin2017real,
  title={Real-time action detection in video surveillance using sub-action descriptor with multi-cnn},
  author={Jin, Cheng-Bin and Li, Shengzhe and Kim, Hakil},
  journal={arXiv preprint arXiv:1710.03383},
  year={2017}
}

License

Copyright (c) 2018 Cheng-Bin Jin. Contact me for commercial use (or rather any use that is not academic research) (email: sbkim0407@gmail.com). Free for research use, as long as proper attribution is given and this copyright notice is retained.

ChengBinJin / ActionViewer

readme