CatchFatCriminals

How it all began

We were out touring in Europe in Napoli when our bags were almost being stolen. The experience was traumatic and we decided no one else should feel that sense of helplessness too. As such, we decided to make CatchFatCriminal.

What it does

CatchFatCriminal uses machine learning to predict from an individual information (Location, Education,Salary) the probability of him being a criminal. After predicting the probability of him being a criminal, the CatchFatCriminal then uses facial recognition software to match an individual to our exciting database.

How we built it

We utilised existing machine learning libraries like scikit-learn and train a criminal prediction decision tree model using data from Kaggle.

For the facial recognition, we utilised the popular dlib library and tensorflow mobilenet implementation to first isolate the human and then match the face to our database, predicting if they are a criminal

Challenges we ran into

Accomplishments that we're proud of

What we learned

We learnt how to prototype rapidly, and how to brainstorm a solution effectively. We also learnt how to split our work between our team, for example, doing the machine learning model, writing the back end server and writing the front end user interface. We learn many technical skills too, such as a Decision tree classifier, Flask in Python, Javascript and HTML, Css

What's next for CatchFatCriminals

We hope to furthur develop the project, by including a way to follow the likely criminal. Currently, even if we do find people who are highly likely to be criminal, there would not be any way to approach him. Perhaps we could include a rover with an attatched camera, to follow the likely criminal and see where they live.

Built With

• dblib
• express.js
• html5
• machine-learning
• node.js
• python
• scikit-learn
• tensorflow