Datasets:
gremlin97
/
RemoteSensingCorpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
820
systematic fairness evaluation when models are transferred
across domains. We examine the performance of two unsu-
pervised domain adaptation methods on a large-scale public
satellite imagery dataset. Model fairness is evaluated be-
tween rural and urban locations as the models are trained
and tested across administrative districts. Based on the ex-
periments, we conclude that the domain adaptation meth-
ods we study can be improved in terms of retaining model
fairness across rural and urban data. Domain adaptation im-
proves overall accuracy at the cost of decreasing fairness on
test domain. Further, more shifts in the raw image distri-
bution and pixel-wise class distribution result in more per-
formance drop. Broadly, our findings demonstrate potential
fairness problems when working with satellite image data
sourced from different locations. Also, the findings indicate
the need for developing methods focused on fair transferlearning, such as through new model architectures or loss
functions.
References
[1] Nabeel Abdur Rehman, Umar Saif, and Rumi Chunara. Deep
landscape features for improving vector-borne disease pre-
diction. In Proceedings of the IEEE/CVF Conference on
Computer Vision and Pattern Recognition Workshops , pages
44–51, 2019. 1
[2] Michiel Bakker, Humberto River ´on Vald ´es, Patrick D Tu,
Krishna P Gummadi, Kush R Varshney, Adrian Weller, and
AS Pentland. Fair enough: Improving fairness in budget-
constrained decision making using confidence thresholds.
2020. 2
[3] Ritika Brahmadesam and Kush R Varshney. Fairly estimat-
ing socioeconomic status under costly feature acquisition,
2021. 2
[4] Joy Buolamwini and Timnit Gebru. Gender shades: Inter-
sectional accuracy disparities in commercial gender classifi-
cation. In Sorelle A. Friedler and Christo Wilson, editors,
Proceedings of the 1st Conference on Fairness, Accountabil-
ity and Transparency , volume 81 of Proceedings of Machine
Learning Research , pages 77–91. PMLR, 23–24 Feb 2018.
2
[5] AP Carleer, Olivier Debeir, and El ´eonore Wolff. Assess-
ment of very high spatial resolution satellite image segmen-
tations. Photogrammetric Engineering & Remote Sensing ,
71(11):1285–1294, 2005. 1
[6] Javiera Castillo-Navarro, Bertrand Le Saux, Alexandre
Boulch, Nicolas Audebert, and S ´ebastien Lef `evre. Semi-
supervised semantic segmentation in earth observation: The
minifrance suite, dataset analysis and multi-task network
study. Machine Learning , pages 1–36, 2021. 2
[7] Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian
Schroff, and Hartwig Adam. Encoder-decoder with atrous
separable convolution for semantic image segmentation. In
Proceedings of the European conference on computer vision
(ECCV) , pages 801–818, 2018. 4
[8] Amanda Coston, Karthikeyan Natesan Ramamurthy, Dennis
Wei, Kush R Varshney, Skyler Speakman, Zairah Mustahsan,
and Supriyo Chakraborty. Fair transfer learning with missing
protected attributes. In Proceedings of the 2019 AAAI/ACM
Conference on AI, Ethics, and Society , pages 91–98, 2019. 3
[9] Ilke Demir, Krzysztof Koperski, David Lindenbaum, Guan
Pang, Jing Huang, Saikat Basu, Forest Hughes, Devis Tuia,
and Ramesh Raskar. Deepglobe 2018: A challenge to parse
the earth through satellite images. In Proceedings of the
IEEE Conference on Computer Vision and Pattern Recog-
nition Workshops , pages 172–181, 2018. 3
[10] Wei Du and Xintao Wu. Fair and Robust Classification Un-
der Sample Selection Bias , page 2999–3003. Association for
Computing Machinery, New York, NY , USA, 2021. 3
[11] Mark Everingham, SM Eslami, Luc Van Gool, Christo-
pher KI Williams, John Winn, and Andrew Zisserman. The
2923
pascal visual object classes challenge: A retrospective. Inter-
national journal of computer vision , 111(1):98–136, 2015. 4
[12] Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pas-
cal Germain, Hugo Larochelle, Franc ¸ois Laviolette, Mario
Marchand, and Victor Lempitsky. Domain-adversarial train-
ing of neural networks. The journal of machine learning
research , 17(1):2096–2030, 2016. 7
[13] Alberto Garcia-Garcia, Sergio Orts-Escolano, Sergiu Oprea,
Victor Villena-Martinez, Pablo Martinez-Gonzalez, and Jose
Garcia-Rodriguez. A survey on deep learning techniques for
image and video semantic segmentation. Applied Soft Com-
puting , 70:41–65, 2018. 4
[14] Omid Ghorbanzadeh, Alessandro Crivellari, Pedram
Ghamisi, Hejar Shahabi, and Thomas Blaschke. A com-
prehensive transferability evaluation of u-net and resu-net
for landslide detection from sentinel-2 data (case study
areas from taiwan, china, and japan). Scientific Reports ,
11(1):1–20, 2021. 2
[15] Arthur Gretton, Karsten M Borgwardt, Malte J Rasch, Bern-
hard Sch ¨olkopf, and Alexander Smola. A kernel two-sample
test. The Journal of Machine Learning Research , 13(1):723–
773, 2012. 7
[16] Ananya Gupta, Elisabeth Welburn, Simon Watson, and Hu-
jun Yin. Cnn-based semantic change detection in satellite
imagery. In International Conference on Artificial Neural
Networks , pages 669–684. Springer, 2019. 1
[17] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun.
Deep residual learning for image recognition. In Proceed-
ings of the IEEE conference on computer vision and pattern
recognition , pages 770–778, 2016. 4