Unsupervised Few-Shot Audio Classification

In the few-shot unsupervised classification task, we assume that only unlabeled data is available during the model pre-training phase. This setting contrasts with traditional methods where labeled data is used for pre-training. Unsupervised Few-Shot Audio Classification aims to achieve efficient and accurate audio classification using a small number of unlabeled audio samples through unsupervised learning techniques. This task not only reduces the cost of data annotation but also enhances the model's generalization ability on new categories, making it highly valuable for practical applications.