Fig 1.
Schematic Diagram of the RTAS-Net Architecture.
Fig 2.
Schematic Diagram of the Improved Encoder Structure and Its Components.
Fig 3.
The Structure of the ASPP Module.
Fig 4.
The Structure of the mini-ASPP Module.
Table 1.
Quantitative comparison on the ISPRS Potsdam dataset. The accuracy for each class is presented in the form of IoU (%). The best results are highlighted in bold.
Fig 5.
Qualitative comparison results of different methods on the Potsdam dataset.
Table 2.
Quantitative comparison on the ISPRS Vaihingen dataset. The accuracy for each class is presented in the form of IoU (%). The best results are highlighted in bold.
Fig 6.
Qualitative comparison results of different methods on the Vaihingen dataset.
Table 3.
Performance Comparison of Different Module Combinations (Potsdam).
Fig 7.
Radar Chart of Per-Class IoU for Ablation Models on the Potsdam Dataset.
Table 4.
Location and Sequential Ablation Results for the Potsdam Dataset (V0–V7).
Fig 8.
Visual Comparison of Segmentation Results from Different Models on the Potsdam Dataset (Red Boxes Highlight Key Comparison Areas).
Fig 9.
Typical failure cases on the Potsdam dataset.
Table 5.
Performance Comparison of Different Module Combinations (Vaihingen).
Fig 10.
Radar Chart of Per-Class IoU for Ablation Models on the Vaihingen Dataset.
Table 6.
Location and Sequential Ablation Results for the Vaihingen Dataset (V0–V7).
Fig 11.
Visual Comparison of Segmentation Results from Different Models on the Vaihingen Dataset (Red Boxes Highlight Key Comparison Areas).
Fig 12.
Typical failure cases on the Vaihingen dataset.
Table 7.
Performance Comparison of Different Module Combinations (LoveDA).
Table 8.
Sensitivity to edge-weight λ in the loss function on LoveDA.
Table 9.
Computational efficiency of module-combination models (A–E) and placement/order variants (V0–V7).