Generalized Zero-Shot Learning - Unseen

In the field of computer vision, Generalized Zero-Shot Learning (GZSL) refers to the task where, during the testing phase, the model not only needs to predict seen categories but also accurately recognize unseen categories. The performance of the model in this task is evaluated by calculating the average of the normalized highest prediction scores for the unseen categories. This task aims to enhance the model's generalization ability when facing new categories and has significant application value.