Algorithm Design and Application of Multi-Modal Large Model in Intelligent Sorting of Image Data

Multi-modal large model shows great application potential because it can fuse and process various types of data. In the field of intelligent sorting of image data, traditional methods are often limited by manual operation or simple automatic identification system, and it is difficult to deal with complex and changeable image data, which leads to limited accuracy and efficiency of sorting. In this paper, a multi-modal algorithm for intelligent sorting of image data is designed in detail. This algorithm effectively improves the accuracy and robustness of feature extraction through innovative text and image modal feature extraction methods combined with feature fusion strategy based on attention mechanism. In the part of algorithm implementation, this paper adopts the text feature extraction method based on Transformer and the improved Convolutional Neural Network (CNN) to extract image features, and designs a feature fusion method based on attention mechanism to realize the deep fusion of text and image modal features. Subsequently, a simple fully connected layer is constructed as the decision-making layer, and the weight and bias are optimized by back propagation algorithm to minimize the classification error. In the experimental part, this paper uses a sample data set with $\mathbf{1 0, 0 0 0}$ text descriptions and corresponding images, and verifies the effectiveness of the proposed algorithm by comparing the performance indicators of different categories. The experimental results show that the algorithm has achieved high accuracy and recall in books, electronic products, clothing and other categories, which is significantly better than the traditional sorting methods based on rules, support vector machines (SVM) and CNN. The application algorithm of multi-modal large model in intelligent image data sorting proposed in this paper not only improves the accuracy and efficiency of sorting, but also provides useful reference for the application of multi-modal large model in wider fields.