Zero-Shot Learning
564 papers with code • 18 benchmarks • 29 datasets
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:
Benchmarks
These leaderboards are used to track progress in Zero-Shot Learning
| Trend | Dataset | Best Model | Paper | Code | Compare |
|---|---|---|---|---|---|
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
|||||
|
Libraries
Use these libraries to find Zero-Shot Learning models and implementations
CUB-200-2011
Oxford 102 Flower
PASCAL Context
AwA
SNIPS
AwA2
aPY
TVQA
LSMDC
MIT-States
Subtasks
Most implemented papers
Learning Transferable Visual Models From Natural Language Supervision
openai/CLIP
•
•
State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories.
Language Models are Few-Shot Learners
By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do.
Prototypical Networks for Few-shot Learning
jakesnell/prototypical-networks
•
•
NeurIPS 2017
We propose prototypical networks for the problem of few-shot classification, where a classifier must generalize to new classes not seen in the training set, given only a small number of examples of each new class.
Learning to Compare: Relation Network for Few-Shot Learning
floodsung/LearningToCompare_FSL
•
•
CVPR 2018
Once trained, a RN is able to classify images of new classes by computing relation scores between query images and the few examples of each new class without further updating the network.
Zero-Shot Learning -- A Comprehensive Evaluation of the Good, the Bad and the Ugly
Due to the importance of zero-shot learning, i. e. classifying images where there is a lack of labeled training data, the number of proposed approaches has recently increased steadily.
Learning Deep Representations of Fine-grained Visual Descriptions
hanzhanggit/StackGAN-v2
•
•
CVPR 2016
State-of-the-art methods for zero-shot visual recognition formulate learning as a joint embedding problem of images and side information.
Sampling Matters in Deep Embedding Learning
CompVis/metric-learning-divide-and-conquer
•
•
ICCV 2017
In addition, we show that a simple margin based loss is sufficient to outperform all other loss functions.
Zero-shot User Intent Detection via Capsule Neural Networks
congyingxia/ZeroShotCapsule
•
•
EMNLP 2018
User intent detection plays a critical role in question-answering and dialog systems.
CPM: A Large-scale Generative Chinese Pre-trained Language Model
However, applying GPT-3 to address Chinese NLP tasks is still challenging, as the training corpus of GPT-3 is primarily English, and the parameters are not publicly available.
Finetuned Language Models Are Zero-Shot Learners
google-research/flan
•
•
ICLR 2022
We show that instruction tuning -- finetuning language models on a collection of tasks described via instructions -- substantially improves zero-shot performance on unseen tasks.
