In recent years, an increasing number of medical engineering tasks, such as surgical navigation, pre-operative registration, and surgical robotics, rely on 3D reconstruction techniques. Self-supervised depth estimation has attracted interest in endoscopic scenarios because it does not require ground truth. Most existing methods depend on expanding the size of parameters to improve their performance. There, designing a lightweight self-supervised model that can obtain competitive results is a hot topic. We propose a lightweight network with a tight coupling of convolutional neural network (CNN) and Transformer for depth estimation. Unlike other methods that use CNN and Transformer to extract features separately and then fuse them on the deepest layer, we utilize the modules of CNN and Transformer to extract features at different scales in the encoder. This hierarchical structure leverages the advantages of CNN in texture perception and Transformer in shape extraction. In the same scale of feature extraction, the CNN is used to acquire local features while the Transformer encodes global information. Finally, we add multi-head attention modules to the pose network to improve the accuracy of predicted poses. Experiments demonstrate that our approach obtains comparable results while effectively compressing the model parameters on two datasets.

Accurate damage assessment is critical for reconnaissance and rescue missions. Damage assessment based on unmanned aerial vehicle images is a relevant topic. However, sharing pre- and post-damage knowledge is hindered by the lack of same-viewpoint images in damage assessment for building restoration and reconstruction work. We propose a novel method for generating images based on the same view before and after damage occurs. A point cloud registration method based on serialized matching is presented for spatial transformation with complex damage variations, achieving lower errors and reducing training time. Then, the transformed data are utilized to train a pre-damage differentiable rendering model, generating the undamaged images of buildings from post-damage drone views. Therefore, damage assessment using a large language model (LLM) is conducted by comparing pre- and post-damage images and point clouds. We also construct pre- and post-destruction datasets based on videos. Our method overcomes the limitation of obtaining paired image data from the same perspective, improving the efficiency of evaluation and reducing the difficulty and cost of data collection. We validated the effectiveness of the method through scaling experiments and real explosion scenarios.