Method, client and system for interactive difficult portrait retrieval

Abstract

The invention discloses an interactive difficult portrait retrieval method, client and system, and relates to the technical field of portrait recognition retrieval. The method includes the steps of: acquiring an image, converting the image into an image representation set based on a semantic attribute label; receiving a query input by a user, and obtaining semantic attribute information selected by the user for the query; marking a significant semantic attribute according to the user's selection, according to Whether the image has a salient semantic attribute Classify the aforementioned image representation set into a salient attribute image set and a non-salient attribute image set, classify and sort the salient attribute image set and the non-salient attribute image set respectively, and according to the sorting results Generate an initial candidate ranking queue; obtain the final goal determined by the user through mixed similarity reranking interactive retrieval. The invention has the advantages of reasonable manual workload, high retrieval success rate, fast convergence speed and wide application range, and is especially suitable for the retrieval application of difficult portraits in the public security industry.

Description

technical field [0001] The present invention relates to the technical field of portrait recognition retrieval, in particular to a method, client and system for interactive difficult portrait retrieval. Background technique [0002] The method of portrait retrieval (also known as portrait image retrieval) mainly uses artificial recognition methods and / or portrait recognition technology to retrieve portraits in image data and obtain retrieval results. The key to portrait retrieval technology lies in the content analysis and understanding of portrait pictures, that is, using machine vision to extract semantic information from portrait pictures for retrieval. The semantic information reflects the content of portraits and also constitutes the features on which portrait retrieval is based. The retrieval algorithm is based on The similarity of semantic features is used to sort the retrieval results. For portrait image retrieval, the semantic feature is required to describe the con...

AI Technical Summary

Problems solved by technology

However, on the one hand, due to the difference in the "sensitive attributes" of human vision and machine vision, there is a large "semantic gap" between human eyes and machine vision, which has caused great troubles to image recognition work; on the other hand, the current interactive The traditional retrieval technology is usually an interactive retrieval method based on random candidate selection, which requires a large number of manual operations, greatly increases the number of cycles in medium and large-scale or similar sample sets, and even falls into an unsolvable state

Images