Accuracy (10 unseen classes): 68.77Accuracy (24 unseen classes): 46.12Random Split Accuracy: 50.67
Papers with Code · CC BY-SA 4.0
Zero-Shot Learning
Zero-shot learning (ZSL) is a model's ability to detect classes never seen during training. The condition is that the classes are not known during supervised learning.
Earlier work in zero-shot learning use attributes in a two-step approach to infer unknown classes. In the computer vision context, more recent advances learn mappings from image feature space to semantic space. Other approaches learn non-linear multimodal embeddings. In the modern NLP context, language models can be evaluated on downstream tasks without fine tuning.
Benchmark datasets for zero-shot learning include aPY, AwA, and CUB, among others.
( Image credit: Prototypical Networks for Few shot Learning in PyTorch )
Further readings:
논문
Accuracy (10 unseen classes): 62.69Accuracy (24 unseen classes): 38.70
Accuracy (10 unseen classes): 63.30Accuracy (24 unseen classes): 44.30
Accuracy (10 unseen classes): 59.53Accuracy (24 unseen classes): 35.77Random Split Accuracy: 45.14
Accuracy (10 unseen classes): 74.15Accuracy (24 unseen classes): 65.06Random Split Accuracy: 69.47
Accuracy (10 unseen classes): 65.74Accuracy (24 unseen classes): 45.30Random Split Accuracy: 46.40
Accuracy (10 unseen classes): 79.99Accuracy (24 unseen classes): 59.42Random Split Accuracy: 71.38
Accuracy (10 unseen classes): 71.95Accuracy (24 unseen classes): 52.01
AUC: .906
Harmonic Mean (5 unseen classes): 68.83Harmonic Mean (12 unseen classes): 49.70