This is a matlab implementation of the proposed U-SPDNet. To run our method, the Matlab 2019 (or higher version) software with the deep learning toolbox is required. You can select this toolbox and other necessary computing components according to the system prompts during the installation of this software. After completing the above step, you need to put the files named UAV_COV (we here take the UAV-Human dataset as an example) and sample_for_SPDNet_UAV.mat under the path '.\data\afew'. Then, run spdnet_afew.m to obtain the classification performance of the designed model versus the number of training epochs.