Honda Machine Learning Method for Bird's-Eye View 3D Object Detection

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published April 7th, 2026

Detected April 7th, 2026

Summary

USPTO granted Honda Motor Co., Ltd. Patent US12597277B2 covering a machine learning method for three-dimensional object detection using bird's-eye view images. The patent describes associating annotations with extended areas in bird's-eye view images and training models to identify 3D objects. The patent contains 12 claims and was granted to inventors Yuji Yasui, Hideki Matsunaga, Takashi Matsumoto, and Gakuyo Fujimoto.

View original document View source feed page

What changed

USPTO issued Patent US12597277B2 to Honda Motor Co., Ltd. for a learning method involving bird's-eye view image processing for 3D object detection. The patent covers associating annotations with extended areas in bird's-eye view images and generating trained machine learning models to output 3D objects from such images. CPC classifications span computer vision (G06V), machine learning (G06N), and automotive applications (B60W).

This patent grant gives Honda exclusive rights to this specific implementation of bird's-eye view 3D object detection using machine learning. Competitors developing autonomous vehicles, ADAS systems, or similar computer vision applications may need to design around these claims or explore licensing options. The patent is particularly relevant to the autonomous vehicle and advanced driver assistance system industries.

What to do next

Monitor for updates
Review patent claims for infringement risk
Consider licensing if developing similar 3D object detection systems

Source document (simplified)

← USPTO Patent Grants

Learning method, learning device, mobile object control device, mobile object control method, and storage medium

Grant US12597277B2 Kind: B2 Apr 07, 2026

Assignee

HONDA MOTOR CO., LTD.

Inventors

Yuji Yasui, Hideki Matsunaga, Takashi Matsumoto, Gakuyo Fujimoto

Abstract

A learning method includes steps of associating, with an extended area in a bird's-eye view image, an annotation indicating that the extended area is a three-dimensional object; and generating, based on training data in which a bird's-eye view image is associated with an annotated bird's-eye view image obtained by assigning an annotation to a three-dimensional object in the bird's-eye view image, a trained model by learning parameters of a machine learning model so that the trained model receives input of a bird's-eye view image to output a three-dimensional object in the bird's-eye view image.