When I looked at the well-planned roads and strict rules in the USA, I thought why not we? why not in India? Aren't we the fastest developing nation? These thoughts had stuck me a long time back from then but didn't find any ways or opportunities to fix it. I as a startup, when I read about Ford hackathon in the collaboration with the Indore government, I didn't waste time and started building Safety Angel, the smart and Machine Learning enabled local community app. I could win this contest, I hope I would get a chance to meet Indore Govt. officials and work together to take this product further for safer roads.

I just want to make sure that a daughter who is waiting for her dad to return home after work to reach home safely. This inspires me the whole time building this product.

What it does

In one sentence, it uses machine learning, crowdsourced data points, and sets, user feedbacks to improvise and promote 'Street for All' initiative.


  1. Virtual Bike Lanes and Footpaths: Because there are no dedicated lanes for bicycles and pedestrians. With this one can easily implement Street for All Initiative process. Suggest/add Virtual Bike Lanes, Virtual Footpaths, Bicycle Parking & Stop Intersections in a click using 'Suggest Rule' feature. Also get all the local community rules while on your way to your destination on a non-distracting interface.
  2. Get alternative routes suggested based on current live traffic, and machine learning algorithms' output which is based on past road traffic, conditions and accident rates.
  3. Get to know what's the risk factors, non-bikeability factors and live route updates on your route.
  4. Get alerted about the upcoming bump, potholes while anonymously we capture the road conditions sensed by your mobile sensors and send it to our ML servers.
  5. Get alerted about the Virtual Bike Lanes and Virtual Footpath Availablity and stay safe. Also get alerted about the Upcoming Bicycle Parking availability and Speed-Limit Zones.
  6. Vehicle Alert: Get alerted about upcoming pedestrians, bicycles, bikes, cars before they are visible to you and get alerted. This feature uses Wifi P2P protocols which are the same protocol used in Wifi Direct apps like ShareIt. This feature works the fastest among all the tech available and also works offline which makes it more suitable for India.
  7. Let your family and friends could track you using 'Track Me' feature.
  8. Send progressive tickets over Safety Angel community using 'Ticketing' Feature, for those who defaults the road rules. In progressive ticketing, if users are on our community app, users would get the warning for 3 times before being reported to local authorities. This initiative would help the driver to know their mistakes and improve.
  9. Send Smileys Reactions to the driver/rider as a token of gratitude for driver safe or let you the road to overtake.
  10. Crowdsourced SOS: Need help? Call SOS volunteers near you rather than your family, friends, and police who are far away.
  11. Non-Distracting Interface: All the UI interfaces while on ride are non-distracting like vehicle alert are alerted through sounds, bump alerts are shown by changing color on the screen and so on and so forth.

More about crowdsourced data node

There are multiple data nodes in our system which is feed with crowdsourced data coming from the Safety Angel application.

  1. First crowdsourced data node is the users' location used to find the nearest/quickest helper for Crowdsourced SOS.

a. Users who volunteer to become a Crowdsourced SOS helper/responder have to activate the feature in 'Track'-->'Be an SOS Volunteer'. Once they activate this feature, their location is updated/tracked by our cloud system every 60 seconds[Could be a higher number but for demo purpose, we kept it small]. b. So whenever a victim is under distress s/he could either press SOS button On-Ride/On-Commute Dashboard or victim could press the headphone's button 4 to 6 times vigorously(pressing headphone's button is added here so as to not let the attacker know that vicitim is calling for help and get aggresive) to send out SOS distress signals to our server. [This feature works only when 'Track Me' feature is turned on. ] c. Now our server matches the victims location with the SOS responders location and automatically sends notifications to them saying 'Victim ABC in danger, help him/her'.

  1. Second set of data nodes for crowdsourced data consists of users' feedback and rating for a bus, potholes and bumps detected by the users' devices, current traffic data provided by Google and current traffic data crowdsourced by Safety Angel app users by turning on their location in 'On Ride/On-Commute Dashboard' while on commute, and local rules or constraits like speed limits, school zones etc.. set by local authorities using 'Suggest Rule' option.

-->For Buses a. When user chooses 'Bus' as a mode of transportation in 'Ride' feature, and selects the destination s/he wants to go from his current location. Our system calculates 1 to 3 best routes the user could choose to reach the destination. Here the best routes are calculated using users feedback given to a specific bus on their route, the average delays by a specific bus on their route, the time of commute, traffic data crowdsourced by the users by turning on their location while on commute and the older days traffic data crowdsourced, and risk factors taken into the account.

Also, User could choose the best route depending on the Risk Factors which is shown as a Risk Percentage Meter on the same screen. Which is calculated using risk factors invloved in the route like accident rates data on the available routes obtained from the earlier period, crime rates on the routes, speed ways etc...

User could also see the avg delays and ratings given by the Safety Angel community app users to the upcoming buses on the same screen to plan their commute.

b. Once the user chooses one route and starts the trip, s/he can see the nearby Safe Bus Stop Waiting areas marked the local authorities or communtity moderators using 'Suggest Rule'-->'Mark Safe Bus Stop Waiting Area' option. [Crowdsourced Input for Safer Routes and Risk Factors]

c. User could call for crowdsourced SOS as mentioned above. Rate Buses and Bus Drivers whichever they board-on. [Input for Bus Avg Delays, Bus Rating]

--> Other Mode of Transportation except Bus. a. When the user chooses other transportation mode expect bus in 'Ride' feature, and selects the destination s/he wants to go from his current location. Our system calculates 1 to 3 best routes the user could choose to reach the destination. Here the best routes are calculated using users feedback given to route after finishing their trip on Safety Angel, roads conditions like potholes and bumps automatically detected by our Safety Angel app users while commuting by turning on the On-Ride/On-Commute Dashboard [to simulate a bump/pothole, go to Ride--> Select Bike/Car as the mode of transport-->Select Destination and Route-->Start Driving--> Shake the phone vigoursoly to simulate a bump, you will get a toast saying 'Bump detected'. Now after 60seconds, you could see a circular green-red gradient patch at the location where the bump was detected by your device], accident rates, etc...

b. Once the user chooses one route and starts the trip, s/he can see the nearby Virtual Footpath, School Zones, Speed Limits marked the local authorities or communtity moderators using 'Suggest Rule'-->'Mark Safe Bus Stop Waiting Area' option.

Total Crowdsourced Nodes:

  1. Users Location --> Used for Crowdsourced SOS --> Used to calculate live traffic flow --> Used to calculate best routes

  2. User Location + Bump and Pothole Detectors --> Used to calculate road conditions

  3. Users Feedback for Routes taken --> Used to calculate best routes for upcoming drivers.

  4. Users Feedback for Bus taken --> Used to calculate best routes for bus commuters.

  5. Local Authorities Input for Local Rules --> Used to alert the drivers and commuters about local community rules. --> Used to calculate risk factors and safer routes.

There are many more minor nodes like this, as these are hard to explain on video in shorter time I explained them in the description. Let me know if you have any queries.

How I built it

I built it using the following software and libraries,

  1. TensorFlow for Machine Learning AI
  2. Python for Machine Learning
  3. MongoDB as database
  4. Play Framework as backend
  5. Android, Android Studio, and Android API
  6. Amazon AWS etc...

Challenges I ran into

Building an application which involves machine learning, recommendation model, real-time tracking, real-time vehicle alerting system involves a lot of moving part and has to be developed carefully. I found it bit hard to develop it in the very short time as I started to work on this lately.

Accomplishments that I'm proud of

I built an application which makes sure that the daughter who is waiting for her dad to return home from work reach safely. + Proud as this is *'Made in India with love' for the India. * + Toughest app built in the very short span of time.

What I learned

Before I could start working on the application, I had to go through all the literature and resources provided to understand the intensity of this problem. The first thing I learn the issues and the existing solutions in India and planned an upgrade around in keeping future in the mind.

Technically speaking, I learned how to manage the design, development, and testing when there is very short time to complete. I was trying to develop dashboard for the traffic authorities and I couldn't finish it up due to lack of time. From next time, I will be much careful about it.

What's next for Safety Angel

I own a startup of 6 people here in Bangalore. If I could win this content, I would try to approach Indore govt. with the proposal to work together to get this application to the action. Hope things go well. *Fingers crossed. :) *

Built With

Share this project: