Relative Attribute

Relative attributes, focusing on comparative judgments like "which image is angrier?", are emerging as a powerful tool in computer vision. Research currently explores their use in image generation, where user preferences guide the creation of images with subtly varying attributes, often leveraging preference-based learning and generative adversarial networks. This approach offers advantages over specifying absolute attribute values, enabling finer control and more intuitive interaction with image generation systems. Furthermore, relative attributes are proving useful in robot localization and image retrieval, providing compact and transferable visual descriptors that complement traditional absolute feature representations.