Tell Stories out of Pictures

A picture is worth of a thousand words, yet, for a computer, the story inside a picture is worth of nothing because it is difficult for a computer to understand images. Physical Linguistic (PL) image understanding engine has been developed with the goal to tell such a story to computers. The PL image understanding engine has a layered structure. The top layer in the PL image understanding engine is a physical linguistic network of which each node represents either a computational noun or a computational verb. The bottom layer consists of banks of image detectors, called PL image detectors, for computational nouns and computational verbs. While there are many similar or equivalent implementing proposals for the physical-linguistic-network layer, the PL image detectors are unique technologies developed in Yang's Scientific Research Institute, LLC.

PL Image Detectors are at the core of the PL image understanding engine and in fact they play the critical role to distinguish PL image understanding engine from other image understanding platform. There are the following three kinds of PL Image Detectors.
•    Type I PL image detectors. The inputs of this kind of image detector are raw images, the outputs can be either a computational noun or a computational verb. The majority of type-I PL image detectors are basic (might not primary) cognitive features such as mouths and lips.
•    Type II PL image detectors. The inputs of this kind of image detector are the outputs of Type-I detectors, the outputs can be either a computational noun or a computational verb. The majority of type-II PL image detectors are composed cognitive features such as faces and human figures.
•    Type III PL image detectors. The inputs of this kind of image detector are the outputs of Type-I and Type-II detectors, the outputs can be either a computational noun or a computational verb. This is a hybrid class mixed by Type-I and Type-II PL image detectors.