Research on Image Semantic Description Method Based on RVC Network

Image semantic description is a natural language description that the computer generates from an image that conforms to the logic of human language, which is an interdisciplinary discipline of computer vision and natural language processing, and has important application prospects in helping visually impaired patients, intelligent driving, and understanding traditional national costumes. Humans are the most basic ability to describe the content of an image in words, but it is difficult for computers to do so. Therefore, as a Xi problem of multimodal studies, image semantic description needs to accurately identify the attributes and relationships between objects in images, and also consider the accuracy and richness of the description sentence syntax. In order to solve the problems of inaccurate description sentences and a large amount of irrelevant information in the process of image semantic description, this study proposes an image semantic description method based on RVC network, which combines Vision Transformer and ResNeXt-101 network to extract image global and local area features, so that the model can extract image features more fully. In order to ensure the accuracy of the image description statement, the model combines the channel attention mechanism to enhance the salient region features of the image and weaken the irrelevant region features, and finally uses BLEU, METEOR, CIDEr, ROUGE and SPICE evaluation indicators to evaluate the results, and the simulation results show that the network can extract the image features well and make the description statements more accurate and rich.