Fig 1.
The overall architecture of CRDFNet.
Fig 2.
Decoder based on Transformer.
Fig 3.
The detailed architecture of the CFAM.
Fig 4.
The detailed architecture of the DAFRM.
Table 1.
Segmentation results of different models on the Potsdam dataset.
Fig 5.
Visualization results of comparative experiments on the Potsdam dataset.
The red arrow points to the misidentified regions.
Table 2.
Segmentation results of different models on the Vaihingen dataset.
Fig 6.
Visualization results of comparative experiments on the Vaihingen dataset.
The red arrow points to the misidentified regions.
Table 3.
Segmentation results of different models on the UAVid dataset.
Fig 7.
Visualization results of comparative experiments on the UAVid dataset.
The white arrow points to the misidentified regions.
Table 4.
Segmentation results of different models on the MSIDBG dataset.
Fig 8.
Visualization results of comparative experiments on the MSIDBG dataset.
Table 5.
Ablation experiments on the Potsdam dataset.
Table 6.
Ablation experiments on the Vaihingen dataset.
Table 7.
Ablation experiments on the UAVid dataset.
Table 8.
Ablation experiments on the MSIDBG dataset.
Fig 9.
Visualization results of ablation experiments on the UAVid dataset.
Fig 10.
Visualization results of ablation experiments on the Potsdam dataset.
Fig 11.
Visualization results of ablation experiments on the MSIDBG dataset.
Fig 12.
Visualization results of ablation experiments on the Vaihingen dataset.
Table 9.
Ablation study on the weight α mIoU (%) was measured for the Potsdam, Vaihingen, UAVid and MSIDBG dataset.
Table 10.
FLOPs, Params, and mIoU of each encoder-based segmentation model on the Potsdam dataset.
Table 11.
Computational complexity analysis on a single NVIDIA GeForce RTX3090 GPU.