A cross-modal pedestrian re-identification method, electronic device, and medium

Abstract

The application discloses a cross-modal pedestrian re-identification method, an electronic device and a medium, and comprises the following steps: acquiring a pedestrian picture dataset and dividing the pedestrian picture dataset into a training set and a test set, and performing data enhancement on the training set to obtain a training set with RGB, IR and Aux modalities; performing feature extraction on pictures corresponding to each modality to obtain class centers corresponding to each modality and modality-independent class centers; inputting pictures in the training set into a cross-modal pedestrian re-identification network in batches to extract deep features; training the cross-modal pedestrian re-identification network, calculating a loss function, and stopping until the network converges; the loss function is the sum of the contrast loss of each batch of extracted deep features and the class centers corresponding to the RGB, IR and Aux modalities and the modality-independent class centers, the sum of the cross-modal mutual information constraint of the RGB / IR and the cross-modal mutual information constraint of the IR / Aux, and the metric loss constraint of the deep features; and the cross-modal pedestrian re-identification network obtained by training is used for cross-modal pedestrian re-identification on the test set.

Images