The average time that it takes for 911 to respond to accidents is 9 minutes and 35 seconds. For every one minute shaved off this, we can save over 10,000 lives.

What it does

A real-time accident response system that is installed within the car and all cameras across the US. This system will use machine learning to detect if there's a collision/crash within the video frame and than based on which system, it will act differently. Possibility to detect the speed of the crash and the impact and showcase the difference. The moment there's a crash: the footage, the user profile, the GPS location, and the impact of the crash are all sent to the nearest hospital and 911 correspondents. Then based on a framework, we can set the priority levels of the crashes: this data can be gathered from the speed of the car, the accelerometer data, how the crash looks etc. and then show a dashboard with the priority levels of each crash and what needs to be tackled first (because the largest problem today is the transferring of calls and waiting time and sometimes time is spent on the most useless calls of 911).

How we built it

We trained a machine learning model to detect car accidents from publicly available car crash footage data. We then trained each model differently for different priority levels. The moment an accident is detected, we send all this data to the emergency and police service. We then built out a web interface for the emergency facilities to track and monitor accidents and deploy the required forces. We can start using the 30 million public cameras available to help do these detections. This helps make accident response to less than 10 seconds from 10 minutes.

Challenges we ran into

TensorFlow model training overnight was not very easy -- we had to figure out exactly how to train the data to detect crashes. It was difficult to navigate through the process to have a 93% accuracy in detecting these crashes.

Accomplishments that we're proud of

Why is this cool:

  1. 911 takes an average of 10 minutes to respond to calls — this reduces everything to seconds.
  2. Their networks are built on legacy systems that are 60 years old — we’re finally modernizing this industry.
  3. The problem of transferability gets solved (a lot of calls get transferred)
  4. The calls can be used to accommodate more important problems that need someone on the other end.
  5. This can be used to file car insurance claims — its a 300 billion dollar industry. Insurance claimers have the hardest, most boring process ever. It is still using snail mail in some companies actually — look at this tweet by one of my favorite YC alum (show suhail’s tweet).
  6. This product actually saves lives, is gov-tech(RFS), and can become a billion-dollar company.

What we learned

That gov-tech is one of the biggest spaces that is open for disruption and it's not a difficult space to get into.

What's next for

Hopefull make it to YC W2020 ;)

Built With

Share this project: