Inspiration: Person of Interest TV show itself. We thought such an idea could be extended to serve a social and practical purpose in today's cities.
What it does: It is a Law enforcement assistant that aids police officials by providing them with an alerting mechanism to detect major past offenders in possible high risk areas. For eg: Past sex offenders in a children’s playground. It is a civic hack that uses face detection and recognition on live security feed to enable police to tighten security in an area if required.
Let's say a person X with a history of major criminal offenses enters an area where the occupants might be especially at risk, such as children or senior citizens. In this case, based on severity of the past offenses, the police are alerted of the presence of person X in the locality. The police may choose to take appropriate measures such as tightening the security in the area.
Our algorithm categorizes crimes based on their degree of severity and color codes them accordingly.
How we built it: We used
- Python and OpenCV for face detection and recognition.
- SQL database for the backend
- Twilio APIs for Text web services (To provide quick and robust alerts to policemen)
Challenges we ran into
- Achieving high accuracy for face recognition with very limited data.
- Sending MMS using Twilio APIs
- Approaching a sensitive subject
Accomplishments that we're proud of
- We built something for a use case we strongly and genuinely believe in.
- Learnt many concepts of Computer Vision and Machine Learning on the go.
What we learned
Hackathons are super fun!
What's next for Person of Interest
It has a lot of scope for advancement -
- Extend the project to narrow down suspects based on the modus operandi, when there is lack of visual data
- Can also have use cases outside Law and order, e.g.: to support businesses by alerting the manager when a premium customer enters the venture.