Method for reasoning large model picture classification in low-hardware underlying equipment environment

The invention discloses a method for reasoning large model picture classification in a low-hardware underlying equipment environment, and the key points of the technical scheme are that the method comprises the following specific steps: 1, picture feature extraction: extracting features of image natural features and image man-made features in the traffic field by using a visual VIT large model of a multi-modal large model, the method comprises the following steps of: acquiring an image, processing and analyzing information contained in the image, extracting information which is not easily interfered by random factors as features of the image, sorting, extracting high-dimensional feature parameters of the image, having high-dimensional coding capability of 1 * 577 * 1024 through a backbone neural network, totally having 577 channels, and each channel having 1024 hidden layer features, all the weights and parameters are obtained by pre-training the VIT large model, so that the high classification precision and