Whether you are a car enthusiast or not, you have probably already been in front of a magnificent car without being able to identify its model and its characteristics. In a parking lot, you don't have time to wait for the owner of the vehicle to ask questions about its wonder. Today, all you have to do is take two pictures of the vehicle you like, and FindMyCar does the research for you.

You're on the verge of finding that sedan, but in order to do so:

Download the App

Find our App on the App Store & Goople and download it for free!
No needed subscription.




Take pictures...

2 photos are to be taken from the car: one in front and one in profile. To maximize the performance of the application, you can record the vehicle on video.

and make your own opinion!

Discover all the characteristics of the car ! (car brand, model, year of release...)



Magic? No, find out how FindMyCAR works...

Have you ever heard of Artificial Intelligence and Machine Learning? To create FindMyCar, it is mainly about supervised machine learning used by ml5 (a JS library). Let's discover together how FindMyCar processes your pictures and draw conclusions (let's call them outputs)

To start, let's say that the photos you send to the algorithm are "inputs". In fact, one must know that those data are confronted to ml5 algorithms that are going to choose the wanted car among a very large selection (let's say classification). Large datasets with human-derived classifications are used to create models ("MobileNet" for FindMyCar case)that ml5 algorithms will use to determine classifications for future inputs. Finally, after giving your "inputs", ml5 algorithms through complex models based on classification of a large dataset, will find the car you're looking for.
You will also get the precision rate of the ml5 algorithms. How much % do they estimate to have the answer to your query?

Let's take a closer look. Let's imagine that the internal system of our application directly converts your two photos into their "https addresses" (like Glitch would do for example). First of all, ml5 must be able to establish the model of the car. The front car photo has this selector: img1 and the side car photo has this selector:img2. When the algorithms have turned (let's see the script below ), according to the "MobileNet" model, the variables classification1 and classification2 which return (respectively) the result found for the car model for img1 and img2, it will be necessary to set up "if" loops in the internal HTML database of FindMyCar. If the variables classification1 and classification2 are equal, and the average of the 2 probabilities, probability1 and probability2 is greater than 0.95, then FindMyLab will be able to tell you the car model. Then, with an external API file, the internal system of FindMyLab will fetch the relevant information of the car model established with ml5. Finally, FindMyLab will show you all the information you came to find. If FindMyLab has a doubt, it will tell you the model it estimates with the associated probability. Here is the script (a part of the full JS script for FindMyCar):


If you want more information about how FindMyCar works...