Skip to content

Latest commit

 

History

History
34 lines (28 loc) · 973 Bytes

README.md

File metadata and controls

34 lines (28 loc) · 973 Bytes

VRN

Violence Recognition Network that uses VGG16 network as base and LSTM as one of the top layers.

Requirements

  • tensorflow==2.4.0
  • opencv-python==4.2.0.34
  • h5py==2.10.0

Required Data

To evaluate/train model, you will need to download the required dataset:

If you want to evaluate/train model on different dataset, you will have to create directory structure as below:

├── dataset
│   ├── Fight
│   │   ├── video0.avi
│   │   └── video1.avi
│   │   ... 
│   └── NonFight
│       ├── video0.avi
│       └── video1.avi
│        ...

Training

Training using our settings:

python vrn.py

All the settings like directory paths or hyperparameters you have to set inside vrn.py file