Table 1.
Summary of remote sensing image segmentation methods based on DL.
Fig 1.
EMNet structure.
Table 2.
Structure diagram of MobileNetV2 network.
Fig 2.
Gate structure.
Fig 3.
Comparison of the decoder block of Deeplabv3+ and EMNet.
(a) Deeplabv3+ decoding module per-forms 4-fold bilinear upsampling; (b) EMNet decoding module performs multi-level upsampling.
Fig 4.
Reprinted from [16] under a CC BY license, with permission from [UAVid], original copyright [2020].
Fig 5.
ISPRS Vaihingen dataset example.
Reprinted from [17] under a CC BY license, with permission from [DGPF], original copyright [2010].
Fig 6.
Comparison of segmentation results on UAVid dataset.
(a) The original image; (b) The label; (c) The segmentation result of Dv_Xtion; (d) The segmentation result of Dv_Mnetv2; (e) The segmentation result of BiSeNetV2; (f) The segmentation result of EMNet. Reprinted from [16] under a CC BY license, with permission from [UAVid], original copyright [2020].
Table 3.
Segmentation performance of different models on the UAVid test set.
Table 4.
mIoU values for each model for different categories on the UAVid test set.
Fig 7.
Comparison of segmentation results on ISPRS Vaihingen dataset.
(a) The original image; (b) The label; (c) The segmentation result of Dv_Xtion; (d) The segmentation result of Dv_Mnetv2; (e) The segmentation result of BiSeNetV2; (f) The segmentation result of EMNet. Reprinted from [17] under a CC BY license, with permission from [DGPF], original copyright [2010].
Table 5.
Segmentation performance of different models on the ISPRS Vaihingen test set.
Fig 8.
Comparison of segmentation results in ablation experiments.
(a) The original image; (b) The label; (c) The segmentation result of Dv_Mnetv2; (d) The segmentation result of Dv_Mnetv2+EDM; (e) The segmentation result of Dv_Mnetv2+MultiL; (f) The segmentation result of EMNet. Reprinted from [16] under a CC BY license, with permission from [UAVid], original copyright [2020].
Table 6.
Ablation analysis of EDM and MultiL on UAVid test set.