In-domain representation learning for remote sensing

By Maxim Neumann, André Susano Pinto, Xiaohua Zhai, Neil Houlsby. 2019

78 citations recorded by the source corpus.

Given the importance of remote sensing, surprisingly little attention has been paid to it by the representation learning community. To address it and to establish baselines and a common evaluation protocol in this domain, we provide simplified access to 5 diverse remote sensing datasets in a standardized form. Specifically, we investigate in-domain representation learning to develop generic remote sensing representations and explore which characteristics are important for a dataset to be a good source for remote sensing representation learning. The established baselines achieve state-of-the-art performance on these datasets.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Media Technology, Representation & Self-Supervised Learning, Remote Sensing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Dacheng Tao
2. Trevor Darrell
3. Kurt Keutzer
4. Bo Li
5. Sergey Levine
6. Stefano Ermon
7. Simone Calderara
8. Chenlin Meng

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Trevor Darrell
3. Dacheng Tao
4. Ming-Hsuan Yang
5. Xiaogang Wang
6. Bernt Schiele
7. Jitendra Malik
8. Andrea Vedaldi

In Media Technology

1. Jocelyn Chanussot
2. Antonio J. Plaza
3. Pedram Ghamisi
4. Danfeng Hong
5. Jon Atli Benediktsson
6. Shutao Li
7. Jing Yao
8. Bing Zhang

In Representation & Self-Supervised Learning

1. Sayna Ebrahimi
2. Trevor Darrell
3. Kurt Keutzer
4. Xiangyu Yue
5. Chenlin Meng
6. Angelo Porrello
7. Stefano Ermon
8. Pau Rodr'iguez L'opez

In Remote Sensing

1. Pedram Ghamisi
2. Jon Atli Benediktsson
3. Jocelyn Chanussot

Citations in the corpus

• A Large-scale Study of Representation Learning with the Visual Task Adaptation Benchmark (2019, neutral): “Further, although the literature reports mixed results (Liu et al., 2017; Raghu et al., 2019; Neumann et al., 2019), we find that they also work for specialized tasks despite the domain-shift (medical or remote sensing).”
• SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery (2022, methodology): “We follow [31, 56] for the train/val splits on EuroSAT.”
• The color out of space: learning self-supervised representations for Earth Observation imagery (2020, methodology): “Instead, [11] proposes indomain fine-tuning, where the pre-training stage performs on a remote sensing dataset.”
• SpectralGPT: Spectral Remote Sensing Foundation Model (2023, methodology): “The training and validation sets align with prior research [40], with 354,196 training samples and 118,065 validation samples.”
• Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data (2021, methodology): “We adopt the same train/val splits proposed in [30].”
• Self-Supervised Pretraining Improves Self-Supervised Pretraining (2021, background): “Model performance generally scales with source dataset size and the similarity between the source and target data [46, 35, 39].”
• Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning (2022, background): “[44] were one of the first to exhaustively share results on existing representation learning and semi-supervised learning techniques for remote sensing imagery.”