A Refined Neural Network Recognition Architecture for Blurred Image Semantic Generalization
Fan Feng, Long Ma · 2020
The technique of generating a caption for a blurred image described by AI still exists the hurdle of recognition. In a blurred image, figuring out a semantic subject is a severe challenge. In this paper, we implement a semantic classifier as an auxiliary oriented filter that combines with a standard dense based caption architecture. This refined architecture is used to categorize a main subject from the media and transform it into a specific predicting range of field. The proposed framework can describe the outcomes and semantical relations that are hidden in an image; they are supposed to deliver a refined natural language sentence through the hierarchical context. Besides, the sample data is evaluated on the proposed approach at the authority benchmark, such as dataset MSCOCO.