# Classification with etalons

In this exercise we will continue with object recognition, and we will implement etalon classification. Etalon is a golden standard, which can represent certain object in our features space. It worth mentioning, that etalon in our feature space, which considers color and shape features, might not be same for different objects. Therefor we need to learn etalons in the learning process.

An etalon is a point that well represents a class of an object. We will compute etalons for each class of objects in our training image. We will do so by summing all elements of particular class, for example all F1 of red stars, and dividing by the total sum of elements, in our case 4. This produce the center of the mass for our feature F1 based on object 1. (For interest, this is standard conditional probability expression).

Where E(j,x )is x coordination of etalon for j class. N is number of feature points and x goes for all objects of j class.

After finding our talons, we are able to classify objects by simply computing geometric distance (in our case Euclidian distance) for our unknown point, from the classification etalons, and therefore we will pick the closes one (winner takes all algorithm).

*Example of etalons*

*Representation of etalons in 2D graph together with normal pieces*

*Text writen to each object based on etalon*

*Sollution in LV*

Data | Value |
---|---|

Source | |

Code |