Datasets:
gremlin97
/
RemoteSensingCorpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
820
support vector machines,” IEEE Geosci. Remote Sens. Lett. ,
vol. 5, no. 3, pp. 336–340, 2008.
[63] W. Kim and M. M. Crawford, Adaptive classification for hyperspec-tral image data using manifold regularization kernel machines,”
IEEE
Trans. Geosci. Remote Sens , vol. 48, no. 1 1, pp. 41 10–4121, 2010.
[64] L. Bruzzone and M. Marconcini, “Domain adaptation problems:
A DASVM classification technique and a circular validation
strategy,” IEEE Trans. Pattern Anal. Mach. Intell. , vol. 32, no. 5,
pp. 770–787, 2010.[65] G. Jun and J. Ghosh, “An efficient active learning algorithm with knowledge transfer for hyperspectral data analysis,” in
Proc.
IEEE Int. Geosci. and Remote Sens. Symp (IGARSS) , Boston,
MA, 2008, pp. I-52–I-55.
[66] S. Rajan, J. Ghosh, and M. M. Crawford, “An active learning
approach to hyperspectral data classification,” IEEE Trans.
Geosci. Remote Sens. , vol. 46, no. 4, pp. 1231–1242, 2008.
[67] G. Matasci, D. Tuia, and M. Kanevski, “SVM-based boosting of ac -
tive learning strategies for efficient domain adaptation,” IEEE J.
Sel. Topics Appl. Earth Observ. , vol. 5, no. 5, pp. 1335–1343, 2012.
[68] B. Settles, Active Learning . San Rafael, CA: Morgan & Claypool,
2012.
[69] D. Tuia, M. Volpi, L. Copa, M. Kanevski, and J. Muñoz-Marí, “A survey of active learning algorithms for supervised remote sens -
ing image classification,”
IEEE J. Select Topics Signal Processing,
vol. 5, no. 3, pp. 606–617, 2011.
[70] M. M. Crawford, D. Tuia, and H. L. Yang, “Active learning: Any value for classification of remotely sensed data?”
Proc. IEEE , vol.
101, no. 3, pp. 593–608, 2013.
[71] L. Bruzzone, C. Persello, and B. Demir, “Active learning methods in classification of remote sensing images,” in
Signal and Im -
age Processing for Remote Sensing , C. Chen, Ed. Boca Raton, FL:
CRC, 2012.
[72] B. Demir, F. Bovolo, and L. Bruzzone, “Detection of land-cover transitions in multitemporal remote sensing images with active
learning based compound classification,”
IEEE Trans. Geosci.
Remote Sens. , vol. 50, no. 5, pp. 1930–1941, 2012.
[73] A. Stumpf, N. Lachiche, J. P. Malet, N. Kerle, and A. Puissant, “Active learning in the spatial-domain for landslide mapping in
remote sensing images,” in
Proc. Euro. Conf. Machine Learn.
(ECML) , Active Learning in Real-World Applications Work -
shop, Bristol, U.K., 2012.
[74] M. Roy, F. Melgani, A. Ghosh, E. Blanzieri, and S. Ghosh, “Land-cover classification of remotely sensed images using compressive
sensing having severe scarcity of labeled patterns,”
IEEE Geosci.
Remote Sens. Lett. , vol. 12, no. 6, pp. 1257–1261, 2015.
[75] N. Alajlan, E. Pasolli, F. Melgani, and A. Franzoso, “Large-scale image classification using active learning,”
IEEE. Geosci. Remote
Sens. Lett. , vol. 11, no. 1, 259–263, 2014.
[76] P. Blanchart, M. Ferecatu, S. Cui, and M. Datcu, “Pattern retrieval in large image databases using multiscale coarse-to-fine cascad -
ed active learning,”
IEEE J. Sel. Topics Appl. Earth Observ. , vol.
7, no. 4, 1127–1141, 2014.
[77] B. Demir, L. Minello, and L. Bruzzone, “Definition of effec -
tive training sets for supervised classification of remote sensing images by a novel cost-sensitive active learning method,”
IEEE
Trans. Geosci. Remote Sens. , vol. 52, pp. 1272–1284, Feb. 2014.
[78] C. Persello, A. Boularis, M. Dalponte, T. Gobakken, E. Naesset,
and B. Schölkopf, “Cost-sensitive active learning with lookahead:
Optimizing field surveys for remote sensing data classification,” IEEE
Trans. Geosci. Remote Sens. , vol. 52, no. 10, pp. 6652–6664, 2014.
[79] T. Luo, K. Kramer, D. B. Goldgof, L. O. Hall, S. Samson, A.
Remsen, and T. Hopkins, “Active learning to recognize mul -
tiple types of plankton,” J. Mach. Learn. Res. , vol. 6, 589–613,
Apr. 2005.
grs
Authorized licensed use limited to: ASU Library. Downloaded on March 08,2024 at 03:13:37 UTC from IEEE Xplore. Restrictions apply.
Towards Geospatial Foundation Models via Continual Pretraining
Mat´ıas Mendieta1*Boran Han2Xingjian Shi3Yi Zhu3Chen Chen1
1Center for Research in Computer Vision, University of Central Florida
2Amazon Web Services3Boson AI
matias.mendieta@ucf.edu boranhan@amazon.com xshiab@connect.ust.hk
yi@boson.ai chen.chen@crcv.ucf.edu
Abstract
Geospatial technologies are becoming increasingly es-
sential in our world for a wide range of applications,
including agriculture, urban planning, and disaster re-
sponse. To help improve the applicability and performance
of deep learning models on these geospatial tasks, vari-
ous works have begun investigating foundation models for
this domain. Researchers have explored two prominent ap-
proaches for introducing such models in geospatial appli-
cations, but both have drawbacks in terms of limited per-
formance benefit or prohibitive training cost. Therefore,
in this work, we propose a novel paradigm for building
highly effective geospatial foundation models with minimal
resource cost and carbon impact. We first construct a com-
pact yet diverse dataset from multiple sources to promote
feature diversity, which we term GeoPile. Then, we in-
vestigate the potential of continual pretraining from large-
scale ImageNet-22k models and propose a multi-objective
continual pretraining paradigm, which leverages the strong
representations of ImageNet while simultaneously provid-
ing the freedom to learn valuable in-domain features. Our
approach outperforms previous state-of-the-art geospatial
pretraining methods in an extensive evaluation on seven
downstream datasets covering various tasks such as change
detection, classification, multi-label classification, seman-
tic segmentation, and super-resolution. Code is available
athttps://github.com/mmendiet/GFM .
1. Introduction
The significance of geospatial technologies has pro-