System and method for domain adaptation using synthetic data
AssigneeSynthesis Ai, Inc.
InventorsSergey Nikolenko, Yashar Behzadi
Life of this patent
Estimating the term…
Abstract
A system and method are disclosed for training a system or a model using any combination of synthetic data and real data. The various aspects of the invention include generation of data that is used to supplement or augment real data, wherein the subject of the data can be segmented. Labels or attributes are automatically added to generated data. The generated data is synthetic data that is created using the seed or real data as well as from other synthetic data. Using the synthetic data, various domain adaptation models can be used and trained for unsupervised domain adaptation.
Claims6
A method for synthetic domain to real domain adaptation, the method comprising:
- identifying a synthetic domain as a source with a synthetic dataset including labeled synthetic images;
- identifying a real domain as a target with a real dataset including unlabeled images;
- identifying at least one common feature between the synthetic domain and the real domain;
- identifying a generative model, using domain adaptive frameworks with direct image-to-image translation, trained using the synthetic domain with the synthetic dataset;
- and
- using the model to apply the common feature in the labeled images of the synthetic dataset to label the images in the real domain having the common feature.
Dependencies are inferred from the claim text.