IMAGE PROCESSING METHOD AND APPARATUS, DEVICE, MEDIUM, AND PROGRAM PRODUCT

原文（中文）

Embodiments of this application disclose an image processing method and apparatus, a device, and a medium. The method includes: obtaining a target image to be processed; inputting the target image to a pre-trained image-text model, a model loss of the image-text model including an image loss, and the image loss being constructed according to a first sample image and a second sample image that is obtained by converting a first sample text configured for describing the first sample image; and obtaining a target text configured for describing the target image and generated by the image-text model. In technical solutions of the embodiments of this application, the generated target text can describe the target image as accurately as possible, thereby ensuring accuracy of the target text.