Abstract: We develop a transductive meta-learning method that uses unlabelled instances to improve few-shot image classification performance. Our approach combines a regularized ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback