Patent · US Active

Grouped mathematical differentiable NMS for object detection

US12354383B2 · kind B2 · utility

0Cited by
1References
19Claims
0Family size

Assignee

Inventors

Key dates

Filing dateFeb 22, 2022
Grant dateJul 8, 2025
Priority date
Expiry dateApr 25, 2044

Classification

  • Technology area (CPC G)Physics
  • CPC primaryG06V10/82
  • WIPO fieldComputer technology
  • WIPO sectorElectrical engineering

Abstract

An end-to-end trainable grouped mathematically differentiable non-maximal suppression (NMS) technique is presented for monocular 3D object detection. First, formulate NMS as a matrix operation and then group and mask the boxes in an unsupervised manner to obtain a simple closed-form expression of the NMS. This technique addresses the mismatch between training and inference pipelines and, therefore, forces the network to select the best 3D box in a differentiable manner. As a result, the proposed technique achieves state-of-the-art monocular 3D object detection results on the KITTI benchmark dataset performing comparably to monocular video-based methods.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.