Implicit Modeling of Object Topology by Guidance with Temporal View Attention

Object recognition developed to the most common approach for detecting arbitrary objects based on their appearance. Statistical pattern analysis methods, are able to extract features from appearing images, enabling a classiflcation of the image content; have reached certain maturity; and achieve excellent recognition on rather complex problems. However, these systems seem not directly scalable to human performance in cognitive sense and appearance does not attribute to understanding the structure of objects. Due to noise, occlusions, and illumination, objects are segmented often poorly with more or less drop outs in contour that yields poor recognition performance, and since object representation enables logical input of spatial arrangements to higher cognitive processes, scene interpretation in cognitive manner gets inhibited. In another paper we proposed the architecture and a simulation of the flve bottom layers of a cognitive vision model by implementing the striate visual cortex as the flrst level. Hence, in this work we focus on the concept of modeling object prototypes from geon recipes on biological formations, such as the circuit of Papez, and show how structure of such formation can be utilized for the modeling of objects. The proposed implementation is exemplifled by an object similar to the Necker cube.