Face attribute editing network based on multi-attention mechanism

The invention provides a face attribute editing network based on a multi-attention mechanism. An additive attention mechanism is introduced into jump connection of a generator, and U-shaped network connection based on the attention mechanism is proposed and used for screening detail features irrelev...

Full description

Saved in:
Bibliographic Details
Main Authors HE YINGXUAN, SU YUKUN, ZHANG KE
Format Patent
LanguageChinese
English
Published 30.10.2020
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
HANDLING RECORD CARRIERS
PHYSICS
PRESENTATION OF DATA
RECOGNITION OF DATA
RECORD CARRIERS
Online AccessGet full text

Cover

More Information
Summary:The invention provides a face attribute editing network based on a multi-attention mechanism. An additive attention mechanism is introduced into jump connection of a generator, and U-shaped network connection based on the attention mechanism is proposed and used for screening detail features irrelevant to attributes, so that the detail features are complementary with decoder features. And meanwhile, a complementary self-attention mechanism is adopted to help the model distinguish an attribute editing area and a non-attribute editing area, so that the attribute editing capability and detail retention capability of the model are ensured. In addition, the complementary self-attention mechanism can increase the modeling capability of long-range dependence in the generated image and increase the geometric constraint of the image. The human face attribute editing network can balance the attribute control capability and detail retention capability of the generative adversarial network, can decouple coherent attribut
Bibliography:Application Number: CN202010583325