[PDF] A Residual-Inception U-Net (RIU-Net) Approach and Comparisons with U-Shaped CNN and Transformer Models for Building Segmentation from High-Resolution Satellite Images | Semantic Scholar (2024)

[PDF] Semantic Reader

DO-UNet is introduced, which strengthens the structure of U-Net and outperforms other models in terms of dice coefficients and accuracy, and visualization results confirm DO-UNet's superiority in accurately segmenting spinal structures.

The combination of optical images and LiDAR point cloud in the deep learning model has increased the prediction performance and provided the integrity of road geometry in woodland and shadowed areas.

According to the compared experiments on the satellite images and public building datasets, the results show that the proposed method has a better performance, compared with other methods based on the same unified hierarchical models, in terms of quantitative and qualitative evaluation.

An adaptive convolutional kernels layer (AKL) is proposed as an architecture that adjusts dynamically to images’ sizes in order to extract comparable spectral information from images of different sizes, improving the features’ spatial resolution without sacrificing the local receptive field for various image applications.

Improved U-Net models, traditional U- net models, FCN models, and EfficientNet models that incorporate channel attention mechanisms are used to identify historic building genes, obtaining the optimal intelligent recognition for historical architectural genes based on deep learning.

It is found that the design of interaction and the choice of immersive technology in virtual surgery development should be closely related to the corresponding interactive operations in the real surgical speciality, which facilitates targeted technological adaptations in the direction of greater applicability and fidelity of simulation.

A novel Offset-Building Model (OBM) is proposed, and a prompt-based evaluation method is introduced, where the model reduces offset errors by 16.6% and improves roof Intersection over Union (IoU) by 10.8% compared to other models.

An efficient model called residual U-Net (RU-Net) is proposed to extract buildings by combining the advantages of U- net, residual learning, atrous spatial pyramid pooling, and focal loss to reduce the parameters and degradation of the network.

A deepened space module is introduced, which can ignore the channels with weak target features and emphasize the effective features, and is embedded in each splicing layer in the upsampling process of U-net to achieve the effect of feature selection.

The successful results of Deeper 1 and Deeper 2 architectures show that deeper architectures can provide better results even if there is not too much data, and Shallower 1 architecture appears to have a performance not far behind deep architectures, with less computational cost, and this shows usefulness for geographic applications.

Compared to available deep learning models, the proposed ARC-Net demonstrates better segmentation performance with less computational costs and is both effective and efficient in automatic building extraction from high-resolution aerial images.

A baseline for model cross-area generalization ability is established and a proposal to fine-tune pretrained models from one area on a small subset of an unseen area, the effectiveness of which depends on the model choice and the data size for tuning is investigated.

The stable encoder– decoder architecture is used, combined with a grid-based attention gate and atrous spatial pyramid pooling module, to capture and restore features progressively and effectively and validate the effectiveness of deep learning in practical scenarios.

Under the direct down-sampling and up-sampled of the inputs and outputs by 4x, experiments demonstrate that the pure Transformer-based U-shaped Encoder-Decoder network outperforms those methods with full Convolution or the combination of transformer and convolution.

A novel Vision Transformer (BuildFormer), with a dual-path structure, designed to encode rich spatial details and a global context path to capture global dependencies, and a window-based linear multi-head self-attention, which strengthens the global context extraction by using large windows and greatly improves the potential of the Vision Trans transformer in processing large-sized remote sensing images.

This work proposes an instruction to assign the perturbation to the intermediate feature representations within the encoder, which considers the spatial resolution of input remote sensing imagery and the mean size of individual buildings in the study area.

...

...