Precious lives are lost because of a simple thing of not changing bad car tires on time. NHTSA estimates that tire failures cause 11000 crashes a year of which many are fatal.Ref( Nowadays technology is available to detect and analyze the image with sophisticated machine learning algorithms. The user can be nudged facilitating his timely tire change and saving precious lives.

The reasons are:

  1. There is no major alert prompting the user to change the tires till it becomes fatal.
  2. Insurance is tied to the bad tires which will motivate the customer to switch tires on time.
  3. The financial part can be subsidized by insurance when linked.

There are systems for monitoring Tire Pressure and mostly focused on gas mileage but not on saving lives.

What it does

The system comprising of raspberry pi camera, load cell with proper lighting is installed on gas stations. Once a car comes to the platform, tire images are captured and sent to AWS cloud and processed by Google & Clarifai machine learning APIs and categorized based on severity of damage. For damaged tires a voice message is broadcasted prompting the driver to tap on the NFC tag which takes him to a web application with dynamically generated unique ID requesting him to fill in the details including insurance.

How we built it

An image recognition system that checks tires at the gas-station First, cameras click pictures of car tires A machine learning algorithm classifies the image based on tire state The system informs you if your tires need to be changed soon Tap on the NFC on the machine, go to the website Make a smart decision

Challenges we ran into

  1. Getting a clear tire image for different illuminations and a setup for the same in a gas station
  2. Identifying bad and good tires using computer vision and machine learning algorithms.
  3. Motivating the user to access the information by making it very readily available using NFC. And tying the random customer to a unique ID that can be tied to an insurance
  4. Prototyping a damaged wheel with 3D printing

Accomplishments that we're proud of

  1. Saving lives by applying technology filling the gaps
  2. An integration of computer vision, machine learning, cloud and IOT solution for a positive impact.
  3. A simulation that can be scaled to real-life scenarios
  4. A machine learning algorithm cannot be trained in a day, so using existing libraries for classifying data
  5. As an additional benefit, gas mileage can be increased by 3.3%

What we learned

It is possible to implement this system with computer vision and machine learning but more training needs to be done for more accurate prediction. Machine learning doesn't happen in a day. It is much easier to use API than before and to integrate such critical applications.

What's next for TireSafe

Systematic training for various tire variations For the system to create a massive positive impact there has to be high resolution vision library built for all kinds of tires in varying degrees of failure. False positives has to be eliminated for conditions like rain,muddy pool etc with multiple sensor input analyzed to zero in on real alerts. All the four tires has to be analyzed to eliminate false positives The system has to implemented on gas stations which is an infrastructure change which can be subsidized by stake holders like car manufacturers, insurance companies, toll collectors for highways and government.

Share this project: