Commodity retrieval method and device and electronic equipment

Abstract

The invention provides a commodity retrieval method and device and electronic equipment, and relates to the technical field of image processing, and the method comprises the steps: firstly, extractingto-be-retrieved features of a to-be-retrieved image based on a pre-trained neural network model; carrying out commodity retrieval based on a pre-established commodity feature library according to to-be-retrieved features of the to-be-retrieved image; wherein the commodity feature library is established by performing feature extraction on a to-be-extracted image after rotation transformation basedon the pre-trained neural network model. The method has the advantages that the problem of high commodity retrieval difficulty in the current new retail scene is solved in two aspects, different images can be distinguished more powerfully, dependence on mass data during model training is reduced, and intelligent development of the new retail industry is promoted on the basis of improving the image retrieval precision.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a product retrieval method, device and electronic equipment. Background technique [0002] The essence of commodity retrieval is image retrieval. Image retrieval is an important research topic in the field of computer vision. In the new retail scene, commodity retrieval is a very critical step. The accuracy of commodity retrieval directly determines the accuracy of retail. model performance. [0003] Currently in the research, most of the retrieved images have large differences in appearance, sufficient light, and no occlusion, so retrieval is not difficult. However, in the product retrieval problem in the retail scene, the main difficulties we face are: the fine-grained characteristics of products, that is, the appearance similarity between products; changes in angles and lighting conditions; occlusion problems, etc. When using the currently proposed image ret...

AI Technical Summary

Problems solved by technology

[0004] In view of this, the object of the present invention is to provide a product retrieval method, device and electronic equipment to alleviate the difficulty of product retrieval in the current new retail scene, to distinguish the differences between different images more effectively, and to reduce the model The dependence on massive data during training, on the basis of improving the accuracy of image retrieval, promotes the intelligent development of the new retail industry

Images