Attribute-based interpretable garment information recommendation method and system, medium and equipment

Abstract

The invention provides an attribute-based interpretable garment information recommendation method and system, a medium and equipment. According to the attribute-based interpretable garment informationrecommendation method, the attribute matching signals are dynamically integrated through an attention mechanism to obtain the matching scores of the garments, so that the information recommendation accuracy is better, meanwhile, the matching signals of the attribute levels are captured through an interaction mechanism, so that the interpretability of garments matching is improved.

Description

technical field [0001] The disclosure belongs to the field of information intelligent processing, and specifically relates to an attribute-based explainable clothing information recommendation method, system, medium and equipment. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] With the vigorous development of the clothing industry, compatibility modeling of complementary clothing plays an important role in many fashion applications such as clothing recommendation and dressing assistant. In recent years, there have been many research works around clothing matching and recommendation. However, existing techniques mainly apply “black-box” deep neural networks to model the compatibility of complementary clothing images and texts, thus inevitably lacking the interpretability of matching results (e.g., why two clothings do not match) . [0...

AI Technical Summary

Problems solved by technology

[0005] However, according to the inventor's understanding, clothing attributes are difficult to obtain in actual recommendation and collocation applications, and if the user is requested to input clothing attributes, the interactive experience of the application will be reduced

Therefore, the current automatic clothing recommendation technology cannot align the visual information and attribute semantics of clothing. The existing datasets for clothing collocation modeling lack fine-grained attribute-level collocation annotations, and it is impossible to learn attribute relationships based on clothing collocation annotations. In addition, for different complementary sets, the same attribute collocation signal may have different confidence levels, how to dynamically allocate the confidence level of the collocation signal between attributes has become an unsolvable problem

Images