The SFID algorithm has been trained with descriptions of numerous images to teach it the different features and combinations of features that make up a dish.
The edge around a pizza is white/beige. In this case there are lots of red areas (tomato sauce), light patches (cheese) and a number of different toppings.
The algorithm defines these features independently using various images where what appears in them has been defined previously.
When the algorithm gets a new image, it calculates numerous features based on the image from the oven compartment (e.g. local edges in the image, small areas containing certain colours, etc.) for the purpose of pattern matching.
These features are combined to create additional, more abstract features. This makes it possible to identify, for example, that a circular area with grated frozen cheese and other toppings is a circular frozen pizza.
