Vehicle attribute identification method based on multi-scale attention model

• Main Authors: WANG WEI, SHI DEWANG, TIAN JIAHAO, CAO JUNLIANG, WANG GANG, SUN GUANGZE
• Format: Patent
• Language: Chinese; English
• Published: 21.10.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a vehicle attribute identification method based on a multi-scale attention model, relates to the field of computer vision, is mainly used in the field of vehicle attribute identification, and mainly comprises five parts, namely a multi-scale feature pyramid network, an attribute positioning module, an attribute classifier, a multi-branch joint optimization strategy and a maximum score voting mechanism. According to the method, a multi-scale feature pyramid structure is introduced, so that the network can adaptively learn vehicle attributes of different scales, then a plurality of attribute positioning modules are introduced into spaces of different scales, features of different levels focus on identification of different vehicle attributes, and the network is optimized by adopting a multi-branch joint training strategy. And processing prediction results of different branches through a maximum voting mechanism to obtain an optimal prediction result. 本发明公开了一种基于多尺度注意力模型的车辆属性识别方法，涉及计算机视觉领域