Transductive Zero Shot Classification

Transductive Zero-Shot Classification is a classification method in the field of computer vision, whose core lies in being able to jointly classify all images in the test set without having to learn from specific category samples beforehand. This method improves the accuracy and robustness of classification by leveraging the overall structural information of the test set, making it particularly suitable for scenarios with scarce data or where new categories are constantly emerging.