State of art in Knowledge representation of image and text pair in AI model

If there is a training data pairs of image and corresponding textual explanation of that image, how to represent this image and corresponding textual meaning as form of knowledge in AI model. So that we don't need much of training data, if I can infuse that knowledge. Idea is to infuse this knowledge and then at inference time, if such image appears then perform action. Ex- If there is a image and respective meaning is 'open the door', then if …

ai ai model art artificial text