In the last few years, the artificial intelligence technology has provided unique methods for design analysis in the art field. With the development of China’s economy and cultural prosperity, graphic design has changed people’s lives greatly. Under the concept of modern scientific and technological creation, the variety, scale, and plasticity of graphic design art are constantly improving, and the diversification of design elements is becoming the development trend. We designed a graphic art element recognition model based on single shot multibox detector (SSD) method through deep learning of visual processing technology. This method can automatically identify various elements in multidimensional graphic art works, thus helping artists and learners analyze an art work better. In this method, we take the SSD structure as the main model backbone and use the improved attention mechanism module feature pyramid transformer to replace the original feature fusion module, inject long-distance dependency into the model, and improve the accuracy of object detection. In addition, we use the public dataset to make the relevant image target detection dataset. Different object detection evaluation metrics are used to evaluate the proposed methods, and several existing methods are selected for comparative experiments. Compared with YOLO V5 object detection model, our method improves 0.53%, 0.67%, 1.33%, and 1.28% on pixel accuracy, mean pixel accuracy, average recall, and mean intersection over union, respectively. The proposed algorithm has a great contribution to the performance improvement of object detection and the auxiliary analysis of multidimensional works of art.