Research on Image Visual Information Storage System Based on Computer Machine Learning

A storage system suitable for SIMT architecture is designed, which mainly includes storage controller, data cache and instruction cache. An automatic image annotation framework based on multi-example and multi-label learning is proposed to express the locality of image features. The image is divided into several visually discontinuous regions by a segmentation algorithm with region constraints. At the same time, feature extraction based on texture and internal structure is carried out. The design uses the integrated Verilog HDL language to realize its hardware circuit, and builds the verification platform based on FPGA to verify the function of the storage system. All aspects of verification show that the designed storage system meets the system requirements.