To streamline the maintenance of existing infrastructures such as bridges, there is growing interest in a method involving the acquisition of the three-dimensional (3D) shape of structures as point cloud data. This enables the construction of new 3D models for maintenance purposes. However, 3D models constructed from point cloud data acquired by LiDAR or Structure from Motion lack essential metadata, including not only geometric information of structural members but also details on damage conditions. Automating the process of inputting such information could enable the efficient creation of 3D models for maintenance. Therefore, this study proposes a method for reflecting information acquired through deep learning-based image analysis into the 3D model of a bridge. In the initial phase of this study, image analysis models were developed utilizing a semantic segmentation model based on deep learning to accurately detect each region within the images. Furthermore, this paper proposes a methodology for integrating information detected from images and reflecting it in the 3D model. Finally, this paper presents a case study, wherein the proposed methodology is applied to an actual bridge. The results demonstrated that integrating multiple image analysis outcomes enables more accurate information reflection than relying on the results from a single image analysis. It is anticipated that the application of this study will enable analyses based on the 3D spatial relationships between member information and associated damage data.