Self-Supervised Prototypical Transfer Learning for Few-Shot Classification
This addresses the challenge of reducing annotation costs in few-shot learning for computer vision applications, representing an incremental improvement over existing unsupervised methods.
The paper tackled the problem of few-shot classification by proposing a self-supervised transfer learning approach that constructs a metric embedding from unlabeled data, which outperforms state-of-the-art unsupervised meta-learning methods on mini-ImageNet and achieves comparable performance to supervised methods with far fewer labels in domain shift settings.
Most approaches in few-shot learning rely on costly annotated data related to the goal task domain during (pre-)training. Recently, unsupervised meta-learning methods have exchanged the annotation requirement for a reduction in few-shot classification performance. Simultaneously, in settings with realistic domain shift, common transfer learning has been shown to outperform supervised meta-learning. Building on these insights and on advances in self-supervised learning, we propose a transfer learning approach which constructs a metric embedding that clusters unlabeled prototypical samples and their augmentations closely together. This pre-trained embedding is a starting point for few-shot classification by summarizing class clusters and fine-tuning. We demonstrate that our self-supervised prototypical transfer learning approach ProtoTransfer outperforms state-of-the-art unsupervised meta-learning methods on few-shot tasks from the mini-ImageNet dataset. In few-shot experiments with domain shift, our approach even has comparable performance to supervised methods, but requires orders of magnitude fewer labels.