Scene graph generation for unlabeled data

This invention describes a way to teach an artificial intelligence (AI) system to understand complex scenes in real-world images, even when those images don't have labels. It works by first training the AI on a set of fake (synthetic) images that are already labeled. Then, it generates new labeled fake images that are designed to look more like the real, unlabeled images. The system uses special techniques to make sure the AI learns to bridge the gap between these different types of images, effectively learning from the unlabeled real data without needing humans to label it.

Why it matters: The rapid advancements in generative AI since 2023, particularly in creating high-fidelity synthetic images and data, could significantly improve the quality of the generated labeled synthetic data, making the domain alignment more effective.

Status: Not in forceListed as no longer active. The specific reason is not in the record we hold.
How hard to build: SpecializedMachine learning software, neural networks, domain adaptation

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10