Action Detection in Smart Surveillance Systems - Web Application

Table of Contents

1. Introduction
2. System Block Diagram
3. Features
   • Logs Page
   • Visualizations Page
4. Demo
5. Screenshots
6. How to Run the Code
7. Authors
8. 🚀 About Us
9. Contact

Introduction

In this project, AI-based smart surveillance system capable of automatically detecting different actions was implemented on the Zewail City video surveillance system. The system reached a recall of 80% and a precision of 70% without fine-tuning using custom data. The system provided different features for the end-user through its graphical interface allowing for easy and convenient data retrieval.

System Block Diagram

Features

Logs page:

1- The user can view the history of actions along with some metadata:

- The *id* of the *camera* that captured that action.
- The *timestamp* indicating the start and end of the action.
- The *Building* and *location* where the action happened.
- The *confidence* of the predicted action.

2- The user can filter by:

- Actions
- Building
- Location 
- Camera 
- Start, end date and time.

Visualizations page:

It provides different visualizations that deliver key insights to the user.

Demo

smart.surveillance.demo.mp4

Screenshots

How to Run the Code

1- Clone the project

  git clone https://github.com/mo-shaffei/action-detection-app

2- Go to the project directory

  cd action-detection-app

3- Install dependencies

• if you have CUDA

    pip install moviepy
    conda install -c pytorch torchvision cudatoolkit=10.2
    conda install -c conda-forge -c fvcore -c iopath fvcore=0.1.4 iopath
    pip install pytorchvideo
    conda install -c anaconda flask
    conda install -c anaconda pymongo
    conda install -c anaconda pandas
    python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'
    pip install scikit-video
    pip install opencv-python

• if you don't have CUDA

    pip install moviepy
    conda install -c pytorch torchvision cpuonly
    conda install -c conda-forge -c fvcore -c iopath fvcore=0.1.4 iopath
    pip install pytorchvideo
    conda install -c anaconda flask
    conda install -c anaconda pymongo
    conda install -c anaconda pandas
    python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'
    pip install scikit-video
    pip install opencv-python

4- Start the server

    flask run

Authors

• @Fatma Moanes

• @Mohamed Elshaffei

• @Mahmoud Ashraf

• @Touka Mohamed

🚀 About Us

We are a group of students studying Communications and Information Engineering at Zewail City of Science and Technology in the final year. This project is our Graduation Project.

Contact

To contact us:

[email protected]

[email protected]

[email protected]

[email protected]