• 

  CARLMAX Logo

• 

  CIM Landing Page

• 

  Backend Flowchart for Plate Detection

• 

  Initial goal of damage detection: Exhibit A

• 

  Initial goal of damage detection: Exhibit B

• GIF

  React UI Demo

• 

  React UI Landing Page

• 

  React UI info input page

• 

  React UI Results + Related + Suggested Cars

ML Powered Social Media E-Commerce App

Overview 💡😂

Selling pre-owned cars is a hassle in this day and age, whether you're thinking of taking up a deal on Craigslist or considering going to a used car dealership in-person. Now, with the introduction of CIM, you can now do this in the comfort of your living room couch--COVID-friendly!

We built a web-based version for simple users who don't want to use the social media feature, and a separate iOS and Android version for people who want to use the social media feature and see what other cars are on the market!

With a simple snapshot, you can now get all the info you need to make smart purchases or sales of any used vehicle. Additionally, we provide a social media platform like Venmo so that you can see what other cars are on the market so you can buy a different car if you'd like! Using labeling and OCR powered by the Google Cloud Vision suite, we are able to instantly provide users with info about a vehicle in question, including:

• Automobile Make, Model, Year, and Submodel
• Current Market Price for used cars of that model
• Major Accident/Salvage History of the specific vehicle
• Issued Recalls and Manufacturer Warnings affecting the vehicle

How we built it 🛠️😳

• Frontend: Web-app takes user input images, pipes them to the backend/vehicle recognition scripts (python), then accepts and displays outputs to the user (vehicle's specifications and price range)
• Backend: Python implementation of the Google CV/OCR suites to detect and interpret license plate/general car image capture; sends feedback to the Edmunds API to retrieve vehicle's specifications and price range

Challenges we ran into 😖🌩️

• Sometimes, OCR wouldn't detect license plate information 100% accurately when presented with test images (namely, the "California" located at the upper part of the plate would not be detected as "California")
• We originally wanted to implement machine learning to detect certain types of damages to the vehicle based on the input image (which would affect the price range to a more specific extent) but it had to be dropped due to time and difficulty implementing it with the front end

Accomplishments that we're proud of 😏👍

We successfully reached the degree of functionality we had envisioned for CIM.

What we learned 🤔💭

Hacking is temporary but memes are forever. Our CARLMAX company chairman, Charles W. Heizer, is not a force to be reckoned with.

What's next for CIM 👴🧓

We would want to implement a social media feature where we can interact with other car sellers to negotiate trades of used cars, and suggest similar cars or cars people have traded before.

Built With

• beautiful-soup
• css
• django
• edmunds
• flask
• google-cloud-vision
• javascript
• python
• react
• requests