Car-ify

Information pulled
EX:1
EX:2

Inspiration

Buying a car is overwhelming, especially for students and first-time buyers. Prices fluctuate wildly, dealerships are inconsistent, and you’re never sure if you’re actually getting a good deal. We wanted something that instantly gives people context — is this car good value right now? Car-ify turns object detection into instant price intelligence.

What it does

Car-ify uses a YOLOv11 based model to detect vehicles in a live feed or uploaded image. Once a car is identified, our system automatically estimates the car’s make, model, year, engine size, horsepower and a one line summary of user reviews. It basically lets you point your camera and gives you a quick review of the car.

How we built it

We trained a custom YOLOv11 model on car datasets to classify make/model visually which then runs backend inference + lightweight API to stream camera frames and return instant results. We have built a very simple interface where pressing the 'L' key gives you a one line summary on the live feed and pressing the 'P' key gives you details on the car in a pre-determined format.

Challenges we ran into

The biggest challenge we ran into was acquiring a database to train the model. The database we used had about 300,000 images and we had to filter through it to only get exterior car images. This reduced our database size to roughly 60,000 images. Another issue with the database was that the images we used are only dated till 2021. The model training required a lot of resources and our laptops were not capable enough of handling that, ultimately we had to trade-off on accuracy so that we could submit a model in time.

Accomplishments that we're proud of

Live image detection that returns meaningful results in real time. This tool helps people without any prior knowledge get a background of the car they are looking at.

What we learned

We learned how much training is required to get accurate results for an image classification project. Model improvement isn't always about accuracy it is about getting the best results possible in what the model returns.

What's next for Car-ify

Add condition grading (good / fair / rough) just from visual cues. Get local postings about the car detected and similar cars from different websites. Mobile integration for easier access.