Pedestrian re-identification learning network construction method and pedestrian re-identification method

• Main Authors: TIE ZHIXIN, WANG WANGMENG, CHEN YANBING, TAO LINGBING
• Format: Patent
• Language: Chinese; English
• Published: 08.08.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a pedestrian re-recognition learning network construction method and a pedestrian re-recognition method, and relates to the technical field of deep learning, and the method comprises the steps: obtaining a pedestrian re-recognition data set, and carrying out the preprocessing of the pedestrian re-recognition data set, and obtaining a training data set; inputting the training data set into attribute soft sharing and a context local network, and extracting global features, local features and attribute features; inputting the global features, the local features and the attribute features into a global identity classifier, a local identity classifier and an attribute classifier respectively to obtain prediction probabilities of pedestrian identities, local identities and pedestrian attributes; calculating the total loss based on the pedestrian identity, the local identity and the prediction probability of the pedestrian attribute; and according to total loss back propagation, continuously