TorchGeo: Deep Learning With Geospatial Data

Stewart, Adam J.; Robinson, Caleb; Corley, Isaac; Ortiz, Anthony; Ferres, Juan M. Lavista; Banerjee, Arindam

doi:10.48550/arxiv.2111.08872

Cited by 6 publications

(7 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This methodology not only excels in mapping cropland boundaries with notable accuracy (86.3%), but also markedly curtails the requisite time and computational resources, establishing itself as a formidable tool for large-scale, high-resolution image analyses. The substantial concordance between the predicted field boundaries and those derived through expert methodologies reaffirms the potential of this tactic, which corroborates the observations documented in previous research commending the proficiency of machine learning algorithms in geospatial analyses [30,31].…”

Section: Discussionsupporting

confidence: 86%

Automated Mapping of Cropland Boundaries Using Deep Neural Networks

Gafurov

2023

AgriEngineering

View full text Add to dashboard Cite

Accurately identifying the boundaries of agricultural land is critical to the effective management of its resources. This includes the determination of property and land rights, the prevention of non-agricultural activities on agricultural land, and the effective management of natural resources. There are various methods for accurate boundary detection, including traditional measurement methods and remote sensing, and the choice of the best method depends on specific objectives and conditions. This paper proposes the use of convolutional neural networks (CNNs) as an efficient and effective tool for the automatic recognition of agricultural land boundaries. The objective of this research paper is to develop an automated method for the recognition of agricultural land boundaries using deep neural networks and Sentinel 2 multispectral imagery. The Buinsky district of the Republic of Tatarstan, Russia, which is known to be an agricultural region, was chosen for this study because of the importance of the accurate detection of its agricultural land boundaries. Linknet, a deep neural network architecture with skip connections between encoder and decoder, was used for semantic segmentation to extract arable land boundaries, and transfer learning using a pre-trained EfficientNetB3 model was used to improve performance. The Linknet + EfficientNetB3 combination for semantic segmentation achieved an accuracy of 86.3% and an f1 measure of 0.924 on the validation sample. The results showed a high degree of agreement between the predicted field boundaries and the expert-validated boundaries. According to the results, the advantages of the method include its speed, scalability, and ability to detect patterns outside the study area. It is planned to improve the method by using different neural network architectures and prior recognized land use classes.

show abstract

Section: Discussionsupporting

confidence: 86%

Automated Mapping of Cropland Boundaries Using Deep Neural Networks

Gafurov

2023

AgriEngineering

View full text Add to dashboard Cite

show abstract

“…Backbone: We use a standard change detection backbone, namely, the FC-Siam-Conc by Daudt et al [57], as implemented by the TorchGeo python package [58]. We adapted it slightly from this implementation, removing the final block (which consists of a 3 × 3 transposed convolution, batch normalization, a ReLU activation and a dropout layer), the dropout from the second to last block, and replacing all that by a 1 × 1 convolutional layer.…”

Section: Setupmentioning

confidence: 99%

Filtering Specialized Change in a Few-Shot Setting

Hermann

Saha

Zhu

2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

The aim of change detection in remote sensing usually is not to find all differences between the observations, but rather only specific types of change, such as urban development, deforestation, or even more specialized categories like roadwork. However, often there are no large public datasets available for very fine-grained tasks, and to collect the amount of training data needed for most supervised learning methods is very costly and often prohibitive. For this reason, we formulate the problem of few-shot filtering, where we are provided with a relatively large change detection dataset and, at test time, a few instances of one particular change type that we try to "filter out" of the learned changes. For example, we might train on data of general urban change, and, given some samples of building construction, aim to only predict instances of these on the test set, all without any explicit labels for buildings in the training data. We further investigate a fine-tuning approach to this problem and assess its performance on a public dataset that we adapt to be used in this novel setting. Code will be available at https://gitlab.lrz.de/ai4eo/cd/-/blob/main/fewShotFilteringCd.

show abstract

“…Remote sensing datasets that are (i) readily available for deep learning applications and (ii) exploiting the spatial, spectral, and temporal dimensions of new satellite missions are still very few. For instance, among the twenty-three benchmark datasets implemented in TorchGeo (Stewart et al, 2021), only two encompass a temporal dimension. There is then an opportunity to build RGB+IR image time series around occurrences spread worldwide.…”

Section: Contextmentioning

confidence: 99%

“…Remote sensing datasets for deep learning applications are currently gaining much interest and are more and more accessible. The very recent launch of TorchGeo (Stewart et al, 2021), a Python library to easily handle geospatial datasets in the PyTorch environment, illustrates the recent and still ongoing progress. However, the available datasets remain currently few and the temporal information provided by satellite revisits is almost never used (Sumbul et al, 2019).…”

Section: Comparison With Other Open Remote Sensing Datasets For Deep ...mentioning

confidence: 99%

“…However, the available datasets remain currently few and the temporal information provided by satellite revisits is almost never used (Sumbul et al, 2019). The available datasets are mostly used for land-cover classification (Helber et al, 2019) or semantic segmentation (Schmitt et al, 2019), as described in the benchmark datasets provided in TorchGeo (see Stewart et al, 2021 of Table 1). Sen12MS is for instance a global dataset including 180,662 patches of Sentinel-1/2 256 x 256 m images and MODIS-derived land cover maps (Schmitt et al, 2019).…”

Section: Comparison With Other Open Remote Sensing Datasets For Deep ...mentioning

confidence: 99%

See 1 more Smart Citation

Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family

et al. 2022

View full text Add to dashboard Cite

Species distribution models (SDMs) are widely used numerical tools that rely on correlations between geolocated presences (and possibly absences) and environmental predictors to model the ecological preferences of species. Recently, SDMs exploiting deep learning and remote sensing images have emerged and have demonstrated high predictive performance. In particular, it has been shown that one of the key advantages of these models (called deep-SDMs) is their ability to capture the spatial structure of the landscape, unlike prior models. In this paper, we examine whether the temporal dimension of remote sensing images can also be exploited by deep-SDMs. Indeed, satellites such as Sentinel-2 are now providing data with a high temporal revisit, and it is likely that the resulting time-series of images contain relevant information about the seasonal variations of the environment and vegetation. To confirm this hypothesis, we built a substantial and original dataset (called DeepOrchidSeries) aimed at modeling the distribution of orchids on a global scale based on Sentinel-2 image time series. It includes around 1 million occurrences of orchids worldwide, each being paired with a 12-month-long time series of high-resolution images (640 x 640 m RGB+IR patches centered on the geolocated observations). This ambitious dataset enabled us to train several deep-SDMs based on convolutional neural networks (CNNs) whose input was extended to include the temporal dimension. To quantify the contribution of the temporal dimension, we designed a novel interpretability methodology based on temporal permutation tests, temporal sampling, and temporal averaging. We show that the predictive performance of the model is greatly increased by the seasonality information contained in the temporal series. In particular, occurrence-poor species and diversity-rich regions are the ones that benefit the most from this improvement, revealing the importance of habitat's temporal dynamics to characterize species distribution.

show abstract

TorchGeo: Deep Learning With Geospatial Data

Cited by 6 publications

References 0 publications

Automated Mapping of Cropland Boundaries Using Deep Neural Networks

Automated Mapping of Cropland Boundaries Using Deep Neural Networks

Filtering Specialized Change in a Few-Shot Setting

Deep Species Distribution Modeling From Sentinel-2 Image Time-Series: A Global Scale Analysis on the Orchid Family

Contact Info

Product

Resources

About