Car recognition using the bag of features method

In this paper, we apply the bag of features method to the car make and model recognition problem. In our implementation of the method, we use the LARS algorithm to optimize the quadratic problem known as the Lasso and by doing so we generate the dictionary of words. That dictionary is then used in conjuction to an image database to obtain a feature vector (each image yields one feature vector), said feature vector is then fed to an supervised classification algorithm (in our case an SVM).