Training and inference using neural networks to predict direction of objects in images

By generating synthetic images through self-supervised learning and generative adversarial networks, and training neural networks using various loss functions, the problems of high training resource requirements and difficulty in obtaining ground truth annotations are solved, achieving efficient recognition of object orientation in images.

CN122391779APending Publication Date: 2026-07-14NVIDIA CORP
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NVIDIA CORP
Filing Date
2020-11-17
Publication Date
2026-07-14

Smart Images

  • Figure CN122391779A_ABST
    Figure CN122391779A_ABST
Patent Text Reader

Abstract

Apparatuses, systems, and techniques for identifying a direction of an object in an image. In at least one embodiment, one or more neural networks are trained to identify a direction of one or more objects based at least in part on one or more features of the object other than the direction of the one or more objects.
Need to check novelty before this filing date? Find Prior Art