Image retrieval method based on sketches

Abstract

The invention discloses an image retrieval method based on sketches. The image retrieval method comprises the following steps of S1, tranining classification models of two CNNs corresponding to the sketches and photos respectively; S2, constructing a retrieval model by using the classification model obtained in the step S1, and training the retrieval model based on quadruplet loss; preprocessing the images in the image library; retrieving a single model; fusing results obtained by the plurality of retrieval models to obtain a final retrieval result. The method is based on the theory that the feature vector spacing corresponding to the sketch and the similar image is reduced, and the feature vector spacing corresponding to the sketch and the heterogeneous image is increased at the same time. Compared with triplet loss, the loss is reduced; quadruplet loss limits the distance between the sketch and the image and pays attention to the heterogeneous spacing of the image at the same time, so that the distribution of different types of images in the final feature space has higher class discrimination, i.e., a larger inter-class distance and a relatively smaller intra-class distance are generated, and thus the retrieval model has better performance.

Description

technical field [0001] The invention relates to the technical field of image retrieval, in particular to an image retrieval method based on a sketch. Background technique [0002] With the popularity of shooting equipment, the improvement of storage device performance and the rapid development of network transmission technology, people can acquire and manage a large amount of image data today. Image retrieval technology can help people find the desired target image quickly and conveniently, but when the target image to be searched is composed of complex scenes and is difficult to describe simply, or the object category label in the image is unclear or unknown, based on text labels or categories The retrieval method is inconvenient to use. The sketch-based image retrieval technology (Sketch Based Image Retrieval, SBIR) can handle this kind of demand very well: only need to draw the sketch without text description, you can query similar target images in the image database. ...

AI Technical Summary

Problems solved by technology

[0003] Sketch-based image retrieval mainly needs to solve the following problems: sketches and photo images have inherent differences in color, background, etc., and sketches also have characteristics such as disproportion of parts of objects, different degrees of simplification and anthropomorphism (such as figure 1 shown), the traditional way of image feature extraction is difficult to solve these problems well

However, the image features output by the model still have large intra-class differences and insufficient inter-class differences, which makes the model have a high degree of fitting on the training data and insufficient generalization ability.

Images