Method for training cross-modal retrieval model, cross-modal retrieval method and related device

Abstract

The invention provides a method for training a cross-modal retrieval model, a cross-modal retrieval method and related devices, relating to the field of artificial intelligence. The method comprises the following steps: determining a reference model by utilizing unsupervised learning; performing knowledge distillation based on the reference model and the training data to obtain similar data of thetraining data; carrying out supervised learning by utilizing the similar data of the training data and the training data, so that a cross-modal retrieval model is obtained. According to the method and the device, relatively high accuracy of the trained cross-modal retrieval model can be ensured under the condition that the label of the training data used for supervised learning does not need to be manually labeled.

Description

technical field [0001] The present application relates to the field of artificial intelligence, and more specifically, to a method for training a cross-modal retrieval model, a method for cross-modal retrieval, and related devices. Background technique [0002] Early retrieval was based on text-to-text retrieval. For example, search engines (such as Google, Bing, etc.) are used to retrieve webpages through keywords. However, with the rapid growth of multimedia services, this kind of retrieval based on text-to-text in the same mode can no longer meet the demand. Cross-modal retrieval has gained increasing attention in the industry. Every kind of multimedia data such as text, pictures, and videos can be regarded as a modal. Using cross-modal retrieval can realize functions such as searching pictures by text, searching text by using pictures, or searching videos by text. [0003] At present, common methods for training cross-modal retrieval models can be divided into superv...

AI Technical Summary

Problems solved by technology

Therefore, labeling large-scale data requires a lot of manpower and time

Therefore, training cross-modal retrieval models with supervised methods is limited in practical applications.

[0005] Although unsupervised methods do not require labeling when training cross-modal retrieval models, the accuracy of cross-modal retrieval models trained by unsupervised methods is usually not as high as that of cross-modal retrieval models trained by supervised methods

Images