Nvidia corporation (20240320993). SCENE GRAPH GENERATION FOR UNLABELED DATA simplified abstract

SCENE GRAPH GENERATION FOR UNLABELED DATA

Organization Name

nvidia corporation

Inventor(s)

Aayush Prakash of Toronto (CA)

Shoubhik Debnath of Sunnyvale CA (US)

Jean-Francois Lafleche of Toronto (CA)

Eric Cameracci of Toronto (CA)

Gavriel State of Toronto (CA)

Marc Teva Law of Ontario (CA)

SCENE GRAPH GENERATION FOR UNLABELED DATA - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240320993 titled 'SCENE GRAPH GENERATION FOR UNLABELED DATA

The abstract discusses approaches for training and using scene graph generators for transfer learning, focusing on reducing discrepancies between latent and output distributions using gradient reversal layers and self-pseudo-statistics.

• Scene graph generation technique can decompose domain gaps into different types of discrepancies.
• Discrepancies related to appearance, label, and prediction can be addressed by aligning latent and output distributions.
• Label discrepancies can be managed using self-pseudo-statistics collected from target data.
• Pseudo statistic-based self-learning and adversarial techniques can handle discrepancies without costly supervision.

• • Potential Applications:**

This technology can be applied in various fields such as computer vision, natural language processing, and robotics for improving transfer learning processes.

• • Problems Solved:**

The technology addresses the challenges of domain gaps and discrepancies in scene graph generation for transfer learning, enhancing model performance and generalization.

• • Benefits:**

The technology improves the efficiency and accuracy of transfer learning models by reducing discrepancies and improving alignment between latent and output distributions.

• • Commercial Applications:**

This technology has commercial applications in industries such as e-commerce, healthcare, and autonomous vehicles for enhancing machine learning models' performance and adaptability.

• • Questions about Scene Graph Generators for Transfer Learning:**

1. How does the technology address discrepancies between latent and output distributions? 2. What are the potential applications of scene graph generators in different industries?

• • Frequently Updated Research:**

Researchers are continuously exploring new methods to improve scene graph generators for transfer learning, focusing on enhancing model robustness and adaptability.

Original Abstract Submitted

approaches are presented for training and using scene graph generators for transfer learning. a scene graph generation technique can decompose a domain gap into individual types of discrepancies, such as may relate to appearance, label, and prediction discrepancies. these discrepancies can be reduced, at least in part, by aligning the corresponding latent and output distributions using one or more gradient reversal layers (grls). label discrepancies can be addressed using self-pseudo-statistics collected from target data. pseudo statistic-based self-learning and adversarial techniques can be used to manage these discrepancies without the need for costly supervision from a real-world dataset.