Crime Detection using Machine-Learning

Crime Detection using Machine Learning and web3 is a project that aims to detect criminal activities in video footage using machine learning techniques and store the information in a database using django and web3 as auth.

Installation

Download or clone the repository

  git clone https://github.com/AkhilAndroid/Crime-Detection-using-Machine-Learning.git
  cd Crime-Detection-using-Machine-Learning-and-web3

Now install the dependencies for web3

  pip install -r requirement.txt
  pip install web3_auth_django-0.7-py3-none-any.whl

If you face any issues while installing web3_auth_django refer this repo

And then install python dependencies

  pip install -r requirement-main.txt

Download the pre-trained models and video from Google Drive.

https://bit.ly/40m9Ka4

Extract the files and place them in the root directory of the project.

To run in Docker

First build the Docker file

sudo docker build -t crime-detection .

Cerify the image was created

docker images

You should see something like this

crime-detection-app          latest     c7b090dc63   3 days ago      1.22GB

You can then run the container by

sudo docker run crime-detection

And then open http://127.0.0.1:8000/api in browser to access the web interface.

Demo

Run the following command

Start the server:

python manage.py runserver

Make sure you install Metamask in

Open http://127.0.0.1:8000/api in browser to access the web interface.

Replace video4.mp4 with your video in main.py

# Load the video
vid = imageio.get_reader('video4.mp4',  'ffmpeg')
cap = cv2.VideoCapture('video4.mp4')

Now run:

python main.py

To check every frame in a video run

python all_frame_check.py

To run multiple video run

python multiple_video.py

Check the result in the website

Please support the development by donating.