Zero Shot Composed Image Retrieval Zs Cir

Zero-Shot Composed Image Retrieval (ZS-CIR) is a subtask in the field of image retrieval, aiming to retrieve target images that are visually similar to a reference image and contain the specific changes indicated in a relative description, without the need for supervised learning. This task enables users to have more precise control over the characteristics of the desired images through bimodal queries, where some features are more accurately conveyed through linguistic descriptions and others are more intuitively expressed through visual means.