Brief

Have you ever felt scared? Are you tired of hearing about people lives being in danger, with nowhere to go for help? Well, we are too that is why we created Danger AI.

With the recent sadening news of Sarah Everard, we felt compelled to do something. Emergency service calls fell by over 26% during the pandemic, whilst crimes such as domestic violence rose extremely quickly. We felt as though this negative correlation is highly dangerous and has left lots of people vulnerable to an attack.

Logistics

How does it work

To use Danger AI, users must navigate to our webpage which allows them to click an alert button to start up backend features. At the backend, the police will gain access to real-time tracking that allows an officer to detect the victim’s location.

Moreover, after the user has hit the SOS alert, the police will also gain access to the user’s whole phone and data. To speed up the process of tracking our application uses machine learning-powered audio recognition to convert surrounding sounds from the phone's microphone to text, which will then be categorized using ML to toxic and non-toxic bubbles. This gives officers a better idea of the situation which can also be used for possible evidence.

The police are also able to access the victim’s camera using AI vision in order to gain a better knowledge of how the situation is being moved. This is done through local police officers tracking the phone through AI vision to take thousands of pictures of the victim's area to detect landmarks. Over time this technology will become better the more the ML is trained with more images.

Step by step

1. Users who are in danger will navigate to our site and press the alert button to alert the police.

2. Once pressed, the police immediately get notified on their server which will come up with a map and an alert.

3. Police receive live tracking via geo cordinates.

4. Speech from the victim's surroundings is collected via their phone microphone and filtered. The red category is for phrases that seem bad and violent which can be used later in court and yellow is for anything suggesting there is no danger.

5. Continous images are taken from the victims camera for the police to understand where they are.

How we built it

We created the backend using Express, Node, mongoDB, Google ML /Maps/Vision/Geocoding. We also used HTML and CSS to develop the front-end website of the application. All of these languages created a complex and highly functional product that provides a secure space for victims to report their real-time attackers without the need of verbally dialing 911.

Our Key Takeaways

Challenges we ran into

We both live in different timezones which made it difficult to communicate with each other at the same time. We also faced challenges such as code not working and layouts not being presented as how we thought they would.

Accomplishments that we're proud of

We are proud to have created software that has a lot of potential to reduce crime and help more victims report their incidences at the instant it happens without having the need to feel like they put themselves in more danger.