Predicting Impervious Land Expansion Using Deep Deconvolutional Neural Networks

Pourmohammadi, Pariya; Adjeroh, Donald A.; Strager, Michael P.

doi:10.1109/igarss.2019.8899234

Cited by 3 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Model Parameters: In the proposed 3-D models, just like 2-D models of deepLandU and deepLandS (see [11]), we use dropout at the final convolutional layer of the encoder, which accelerates the training and prevents overfitting. The decoder has more convolutional layers with more upsamplings.…”

Section: B Model Structurementioning

confidence: 99%

“…This metric is used for assessing the performance of binary classification models. We used AUROC in [11] to measure the model performance; however, we noticed that the AUROC value can be too optimistic about the performance of our models. The reason is the highly imbalanced datasets used in this study, in which an excessive improvement in the number of false positives changes the FPR negligibly.…”

Section: ) Area Under Receiver Operating Curve (Auroc)mentioning

confidence: 99%

“…Our results show how these techniques enhance the model performance in predicting the land cover transformation. The results of previously proposed models of deepLandU, deepLandS, and U-Net [11], [18] are used as the baseline models. A major shortcoming of the baseline models, as suggested in [18], is that these models do not incorporate interchannel relationships.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Ensemble Encoder–Decoder Models for Predicting Land Transformation

Pourmohammadi

Strager

Adjeroh

2021

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

Self Cite

View full text Add to dashboard Cite

Land development is a dynamic and complex process influenced by a system of interconnected driving variables. Predicting such a process is important in mitigating severe climate situations and improving the resiliency of communities. Current predictive models in land transformation have not paid a serious attention to capturing and exploiting the interchannel relationships. Moreover, these models often have problems with generalization, which results in poor performance during testing. In this study, we use a novel multichannel data cube, constructed from socioeconomic attributes, terrain characteristics, and landscape traits, to predict land transformation in a watershed in the US. In particular, we introduce methods for projecting impervious land transformations using these data cubes, using 2-D and 3-D convolutional neural networks (CNNs) and their ensembles. We apply fusion at decision, score, and feature levels to improve the generalization ability and robustness of the proposed predictive models. Performance is assessed using the Dice coefficient, receiver operating characteristic curves, data visualization, and running time. Our study shows that the use of 2-D and 3-D CNN ensembles improved the performance of the models in terms of model stability, precision and recall, and Dice coefficient.

show abstract

Section: B Model Structurementioning

confidence: 99%

Section: ) Area Under Receiver Operating Curve (Auroc)mentioning

confidence: 99%