The term of "finding templates using classifiers"

Question

Template image is a sample of some object, shape or any other pattern represented as a digital image. Imagine you have a set of images, containing multiple geometric shapes. An algorithm might segment each shape, of each image, creating a template for each one. In this case, one example of template of a shape x is the sumbimage in image i1 that it is contained.

For each template, some set of features are extracted. In this case, the features should be the number of sides, the length of each side and the angles between them. Using this features, a classifier could detect an specific shape in a set of images.

Therefore, I classifier could analyze each template and answer whether or not it is related to a square shape (four sides with the same length, 90 degrees angle between them). This is called feature-based matching. Without any classification rule, you might compare the template of two shapes and answer whether or not they are the same. Using a set of samples of the same shape, you can apply machine learning algorithms to train a classifier to automatically learn the rules that classify that specific shape.

In this context, “finding templates using classifiers” may means (i) segment each object/shape/pattern in a image, (ii) creates a template, (iii) extract its features, (iv) classify to determine whether or not it is a new instance of a target object/shape/pattern.