<p>In computer vision, visible-infrared pedestrian re-identification (VI Re-ID) is a challenging job, aiming at recognizing and matching pedestrian images captured by different sensor modalities (e.g., visible and infrared light). According to the methods of predecessors, this article proposes the new feature-enhanced network (F-Enet) to re-identify visible-infrared pedestrians. After the basic ResNet feature extraction, the Lightweight Feature Augmentation Module (LAM) constructed in this paper is introduced compared with the traditional augmentation module, the LAM module is designed to be more lightweight with less computational overhead. This makes it possible to enhance the performance of the model without significantly increasing the computational complexity. The LAM module is able to dynamically adjust the weights of the feature maps to enhance the feature representation. This is particularly useful for handling complex visual tasks. Proposes a novel Comprehensive Multi-Dimensional Attention Mechanism Module (CMAMM), which effectively enhances the model’s adaptive to different localized regions attention ability. In this article, two recognized cross-modal datasets are used for experiments, SYSU-MM01 and REGDB, and the results of the experiments show that this method exhibits desirable performance on both standard metrics.</p>