Image retrieval based on feature fusion and local contrastive learning
Image retrieval is aim at finding out several images that are most similar to the query image from the huge image da-tabase.In recent years,the two-stage image retrieval methods using global feature retrieval and local feature reranking have achieved excellent performance,but the reranking stage leads to the problem of slow response time,and local features are trained mainly by the global classification loss function,this method results in a large amount of redundant information in local features.To solve this problem,this paper introduces a single-stage image retrieval method based on feature fusion and local contrastive learn-ing.The experimental results in image retrieval task show that the image retrieval method proposed in this paper demonstrates strong performance,and provides useful enlightenment for the research and application of image retrieval.
万方数据
