A Review of Zero-Shot Image Classification

Zero-shot image classification is a technique for classifying images when there is no intersection between the training set samples and the test set samples. This technique can solve the problem of missing category labels and is an effective means to cope with the problem of new categories that are emerging and can be recognized effectively. This technique has received a lot of attention from scholars since it was proposed. This paper systematically reviews the development of zero-shot image classification techniques: first, we introduce what zero-shot image classification is and its development history, then we systematically summarize the methods of zero-shot image classification techniques, followed by the common datasets and evaluation criteria in this field, and finally, we analyze the problems encountered in the field of zero-shot learning.