Datasets:
gremlin97
/
RemoteSensingCorpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
820
lected over Vaihingen, Germany at a GSD of 0.9m. We
employ the data split implemented in the MMSegmenta-
tion library [10] for our experiments, with 344 training and
398 for validation, all with an image size of 512x512 pix-
els. The WHU Aerial building [21] dataset is sampled over
Christchurch, New Zealand at a GSD of 0.3m. Image tiles
are provided at 512×512pixels, split into 4736 for training
and 2416 for evaluation.
We report the intersect of union (IoU) segmentation re-
sults for all methods in Table 6. ImageNet pretrained mod-
els are notably strong performers in all cases. On both
datasets, SeCo lags substantially behind its ImageNet coun-
terpart. Interestingly, SatMAE is able to bring improvement
over ImageNet-22k on WHU, but fails to do so to a larger
degree on Vaihingen. However, our approach is able to
leverage the already strong ImageNet-22k representations
and guide them towards the geospatial domain, resulting in
overall improvement.
4.4. Super-resolution
In the previous experiments, we evaluated several com-
mon high-level tasks. Nonetheless, the low-level task of
super-resolution is also important in the geospatial domain.
For this task, we re-purpose the SpaceNet2 dataset, which
contains 10,593 8-band images from four cities: Las Ve-
gas, Paris, Shanghai, and Khartoum. The data is provided
at both a GSD of 1.24m (multi-spectral, 162x162 pixels)
and 0.3m (pan-sharpened multispectral, 650x650 pixels).
We formulate a super-resolution task, taking as input the
1.24m multi-spectral images and generating the 0.3m pan-
sharpened equivalent. We evaluate the super-resolution per-
Table 6. Results on the WHU Aerial and Vaihingen segmentation
datasets. We finetune all methods for 40k iterations, and report the
IoU for the building class on WHU and mean IoU (mIoU) across
the 6 classes (impervious surface, building, low vegetation, tree,
car, clutter) of Vaihingen.
Method WHU Aerial Vaihingen
ResNet50 (ImageNet-1k) [19] 88.5 74.0
SeCo [28] 86.7 68.9
ViT (ImageNet-22k) [14] 81.6 72.6
SatMAE [9] 82.5 70.6
Swin (random) [25] 88.2 67.0
Swin (ImageNet-22k) [25] 90.4 74.7
GFM 90.7 75.3
Table 7. SpaceNet2 Super-resolution Results. Notably, while
SatMAE fails to enhance its baseline (ViT ImageNet-22k), our
method exhibits substantial improvement over its respective base-
line (Swin ImageNet-22k) in both PSNR and SSIM.
Method PSNR ↑SSIM↑
ViT (ImageNet-22k)[14] 23.279 0.619
SatMAE [9] 22.742 0.621
Swin (random) [25] 21.825 0.594
Swin (ImageNet-22k) [25] 21.655 0.612
GFM 22.599 0.638
formance of our model and several baselines with the peak
signal-to-noise ratio (PSNR) and structural similarity in-
dex measure (SSIM) in Table 7. The ViT-L ImageNet-22k
model and our model are among the best in terms of PSNR
and SSIM, respectively. Interestingly, SatMAE is not able
to improve over its baseline. On the other hand, our method
improves considerably over its ImageNet-22k baseline.
5. Ablation Studies
We perform multiple ablation studies on the choice of
distillation stage, student initialization, training objectives,
the pretraining dataset components. Further detailed results
and discussions are provided in the supplementary material .
5.1. Distillation Stage
When implementing our feature map distillation objec-
tive, a natural question is at which point should the map-
ping take place. We experiment with different locations
by stage in the Swin transformer and calculate the corre-
sponding ARP in Figure 5. Overall, performing the dis-
tillation after Stage 3 yields the highest ARP. Hence, we
employ this scheme for all downstream experiments. This
result is also intuitively expected; distilling at Stage 3 gives
a large portion of the model the supervisory signal from the
teacher, while still allowing for purely domain-specific fea-
ture learning in the final layers.
(a) (b)
ARPFigure 5. a) Distillation stage ablation results. b) Student initializa-
tion ablation results. “Both” indicates that the teacher and student
branches are initialized with ImageNet weights prior to geospatial
pretraining. “Teacher” indicates that just the teacher branch is ini-
tialized, as described in Section 3.3.
Table 8. GeoPile pretraining dataset ablation. We remove each
dataset individually from GeoPile and report the number of im-
ages remaining and resulting ARP. The row “w/o curated datasets”
removes all data other than NAIP imagery.
Data # Images ARP ↑
w/o WHU-RSD46 444,061 1.77
w/o MLRSNet 451,793 2.17
w/o Resisc45 529,454 1.57
w/o PatternNet 557,554 1.79
w/o curated datasets 300,000 0.53
w/o NAIP 260,954 1.50
5.2. Student Initialization
In our proposed framework, we maintain the teacher
model frozen with ImageNet pretrained weights, and ran-
domly initialize the student. Another alternative is to initial-
ize the student also with ImageNet weights prior to begin-
ning the geospatial pretraining process. However, as shown