LgoBE

LgoBE / CarMakeandModelRecognition / 0.4.5

README.md

Introduction

This algorithm provides a solution to classify cars by their make, model, body style and model year from images. The capabilities provided so far are as follows:

  • Recognizing over 2400 car models
  • Detection from multiple angles
  • Coverage of a great portion of Western brands and models, from 2000 (and older) to present
  • Providing top 3 most accurate results

The algorithm is trained on European and US cars. We are updating it regularly with newly introduced car models.

Update: Version 0.4 brings a major technology platform (Apache MXNet) and algorithm update, improving the accuracy of the classification substantially.

Example

This photo of a car, taking by a security camera at a parking lot, is correctly identified as the 2003 Daewoo Lacetti Hatchback:

[
  {
    "body_style": "Hatchback",
    "confidence": "0.30",
    "make": "Daewoo",
    "model": "Lacetti",
    "model_year": "2003"
  },  ...
]

Input

The algorithm takes an image URL as input. The image should be in JPEG or PNG format. Note that all images will be resized to 256x256 pixels. Images are downloaded with Algorithmia's SmartImageDownloader utility ( https://algorithmia.com/algorithms/util/SmartImageDownloader ).

"https://upload.wikimedia.org/wikipedia/commons/thumb/9/99/Opel_Meriva_B_front_20100723.jpg/640px-Opel_Meriva_B_front_20100723.jpg"

Output

The algorithm will return a top 3 of car models, ordered by confidence in the prediction.

[
  {
    "body_style": "MPV",
    "confidence": "1.00",
    "make": "Opel",
    "model": "Meriva",
    "model_year": "2010"
  },  ...
]

Make: name of the car brand, e.g. BMW, Ford, Land Rover.

Model: name of the car model, e.g. X5, Mustang, Discovery

Body_style is one of these values:

Convertible, Coupe, Hardtop, Sedan, Wagon, SUV, Cargo Van, Club Cab, Crew Cab, Double Cab, Extended Cab, King Cab, Mega Cab, Quad Cab, Regular Cab, SuperCab Fastback, Hatchback, Van, Wagon Van, Minivan, Passenger Van, MPV, Mini MPV, Buggy, Racing, Roadster, Truck

Model_Year: The model year, or the start-of-production year.

Confidence: a number between 0 and 1 indicating the confidence in the prediction, rounded to two decimal places.