SafetyNet

SafetyNet is an implementation of Bi-LSTM RNN to process live feed from webcam to detect violence.
It is aimed to be implemented with police bodycams to help prevent accidental loss of lives due to misinterpretation of actions by law enforcement officials.
There is also a deployment version of SafetyNet on Huggingface Spaces, which can be used as a simple Threat Detection System over pre-recorded mp4/avi/webm videos. Visit the Space here.

Set Up

1. Download the Real Life Violence Situations Dataset from Kaggle. Download link given in requirements.txt.
2. Unzip the dataset and move it to ASSETS folder.
3. Open RLVDetection_Model.ipynb notebook to train and save the model.
4. To load and test the model with real-time footage, use SafetyNet.ipynb.