Patent · US Active

Edge computing

US10726302B2 · kind B2 · utility

3Cited by
0References
20Claims
0Family size

Assignee

Inventors

Key dates

Filing dateNov 29, 2018
Grant dateJul 28, 2020
Priority date
Expiry dateMar 21, 2039

Classification

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

Abstract

Methods, systems, and devices for object localization and classification are described. A device may configure a first unit of a detection layer associated with a learning framework when a quantity of output feature channels of an input feature maps is less than or equal to a quantity of input feature channels of the input feature maps. The first set of layers may include a group convolution layer, a pointwise layer, a batch normalization layer, or a rectified linear layer, or a combination thereof. The device may also configure, a second unit of the detection layer associated with the learning framework, when a second quantity of output feature channels of the input feature maps is less than or equal to a second quantity of input feature channels of the input feature maps. The second set of layers may include a depthwise layer or a pointwise layer, or both.

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