Smart City CCTV Violence Detection Project

Dataset

SCVD (Smart City CCTV Violence Detection) Dataset

System Requirements

Python Setup

• Python Version: 3.10
• Virtual Environment Setup:

  $env:Path = "C:\Users\srsho\AppData\Local\Programs\Python\Python310;" + $env:Path
  $env:Path = "C:\Users\srsho\AppData\Local\Programs\Python\Python310\Scripts;" + $env:Path
  
  python --version
  ./venv/Scripts/Activate

Local Runtime (Google Colab with Docker)

• Supported Models:
  • C3D
  • ResNet

Dependencies

• opencv-python
• matplotlib
• Keras-Preprocessing
• scikit-image
• tensorflow
• pandas
• torch

Common Troubleshooting

PyTorch Device Compatibility Error

Error:

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

Solution: Ensure model and input are on the same device:

model.to(device)

Model Specifications

VideoResnet2

• Training Resolution: 112x112
• Frame Rate: 16 fps

Installation

1. Clone the repository
2. Set up virtual environment
3. Install dependencies:

   pip install -r requirements.txt