Deep learning-based brand LOGO semantic structure automatic generation method

Abstract

The invention discloses a brand LOGO semantic structure automatic generation method based on deep learning, and the method comprises the following steps: S1, collecting brand semantic input information, and constructing a semantic inducement particle set; s2, initializing a composition reaction kernel in a two-dimensional graph incubation space, and constructing a primitive initial state set; s3, constructing a primitive behavior modeling network, and generating a primitive behavior strategy vector; s4, according to the primitive behavior strategy vector, generating a new primitive and updating the primitive initial state set; s5, calculating semantic response saturation in each region, and terminating region primitive growth when the semantic response saturation exceeds a set threshold value; s6, based on all generated primitives and corresponding semantic paths, constructing a LOGO semantic structure map; and S7, performing consistency detection on the LOGO semantic structure atlas, and outputting the corrected LOGO semantic structure atlas, the corrected LOGO sketch and the corrected semantic interpretation report. According to the method, semantic particle modeling and primitive behavior modeling networks are fused, and automatic generation of the LOGO semantic structure is achieved.

Images