Zero Shot Learning

Encord Computer Vision Glossary

Zero-Shot Learning

Zero-shot learning is a machine learning technique in which a model is trained to recognize and classify new objects without explicitly being trained on those objects' examples. This is achieved using a knowledge transfer approach, where the model is given information about the relationships between classes and can use that information to make predictions for previously unseen objects.

From scaling to enhancing your model development with data-driven insights
medical banner

In zero-shot learning, the model is typically trained on a large labeled dataset that includes examples from a set of known classes, as well as additional information about the semantic relationships between those classes. This additional information may include textual descriptions, attributes, or relationships to other classes. The model is then able to use this information to predict the class labels of previously unseen objects that belong to new, unseen classes.

One of the advantages of zero-shot learning is that it allows for more flexible and adaptive models, as they can generalize to new classes without requiring additional labeled data. This makes it particularly useful in scenarios where new classes of objects may emerge over time, or where labeled data is scarce or expensive to obtain.

Zero-shot learning has been applied in a variety of fields, including natural language processing, computer vision, and speech recognition. For example, it can be used to train models to recognize new types of objects in images or to generate natural language descriptions of previously unseen objects.

Although zero-shot learning is still an active area of research, it shows great potential for developing more intelligent and adaptive models that can learn from diverse sources of information and generalize to new tasks with limited or no labeled data.

cta banner

Discuss this blog on Slack

Join the Encord Developers community to discuss the latest in computer vision, machine learning, and data-centric AI

Join the community
cta banner

Automate 97% of your annotation tasks with 99% accuracy