Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates
Recent small-scale studies for pixel-wise labeling of potential landslide areas in remotely-sensed images using deep learning (DL) showed potential but were based on data from very small, homogeneous regions with unproven model transferability. In this paper we consider a more realistic and practica...
Main Authors: | , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2022-01-01
|
Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9780028/ |
_version_ | 1797990089395011584 |
---|---|
author | Savinay Nagendra Daniel Kifer Benjamin Mirus Te Pei Kathryn Lawson Srikanth Banagere Manjunatha Weixin Li Hien Nguyen Tong Qiu Sarah Tran Chaopeng Shen |
author_facet | Savinay Nagendra Daniel Kifer Benjamin Mirus Te Pei Kathryn Lawson Srikanth Banagere Manjunatha Weixin Li Hien Nguyen Tong Qiu Sarah Tran Chaopeng Shen |
author_sort | Savinay Nagendra |
collection | DOAJ |
description | Recent small-scale studies for pixel-wise labeling of potential landslide areas in remotely-sensed images using deep learning (DL) showed potential but were based on data from very small, homogeneous regions with unproven model transferability. In this paper we consider a more realistic and practical setting for large-scale heterogeneous landslide data collection and DL-based labeling. In this setting, remotely sensed images are collected sequentially in temporal batches, where each batch focuses on images from a particular ecoregion, but different batches can focus on different ecoregions with distinct landscape characteristics. For such a scenario, we study the following questions: (1) How well do DL models trained in homogeneous regions perform when they are transferred to different ecoregions? (2) Does increasing the spatial coverage in the data improve model performance in a given ecoregion? and (3) Can a landslide pixel labeling model be incrementally updated with new data, but without access to the old data and without losing performance on the old data? We address these questions by developing a mechanism for incremental training of semantic segmentation models. We call the resulting extension task-specific model updates (TSMU). A national compilation of landslide inventories by the U.S. Geological Survey (USGS) was used to develop a global database for this study. Our results indicate that the TSMU framework can be used to aid in the creation of new landslide inventories or expanding existing datasets, and also to rapidly develop hazard maps for situational awareness following a widespread landslide event. |
first_indexed | 2024-04-11T08:29:45Z |
format | Article |
id | doaj.art-2d90ab8149a54d9eb8bdeed86d342002 |
institution | Directory Open Access Journal |
issn | 2151-1535 |
language | English |
last_indexed | 2024-04-11T08:29:45Z |
publishDate | 2022-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
spelling | doaj.art-2d90ab8149a54d9eb8bdeed86d3420022022-12-22T04:34:33ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352022-01-01154349437010.1109/JSTARS.2022.31770259780028Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model UpdatesSavinay Nagendra0https://orcid.org/0000-0003-4350-0307Daniel Kifer1Benjamin Mirus2https://orcid.org/0000-0001-5550-014XTe Pei3https://orcid.org/0000-0002-2154-8505Kathryn Lawson4Srikanth Banagere Manjunatha5https://orcid.org/0000-0003-4994-8022Weixin Li6Hien Nguyen7https://orcid.org/0000-0002-7237-4752Tong Qiu8https://orcid.org/0000-0003-2516-6851Sarah Tran9Chaopeng Shen10Department of Computer Science and Engineering, Pennsylvania State University, University Park, PA, USADepartment of Computer Science and Engineering, Pennsylvania State University, University Park, PA, USALandslide Hazards Program, U.S. Geological Survey, Golden, CO, USADepartment of Civil and Environmental Engineering, Pennsylvania State University, University Park, PA, USADepartment of Civil and Environmental Engineering, Pennsylvania State University, University Park, PA, USADepartment of Computer Science and Engineering, Pennsylvania State University, University Park, PA, USADepartment of Civil and Environmental Engineering, Pennsylvania State University, University Park, PA, USASchool of Science, Engineering, and Technology, Pennsylvania State University—Harrisburg, Harrisburg, PA, USADepartment of Civil and Environmental Engineering, Pennsylvania State University, University Park, PA, USAGoogle Inc., Mountain View, CA, USADepartment of Civil and Environmental Engineering, Pennsylvania State University, University Park, PA, USARecent small-scale studies for pixel-wise labeling of potential landslide areas in remotely-sensed images using deep learning (DL) showed potential but were based on data from very small, homogeneous regions with unproven model transferability. In this paper we consider a more realistic and practical setting for large-scale heterogeneous landslide data collection and DL-based labeling. In this setting, remotely sensed images are collected sequentially in temporal batches, where each batch focuses on images from a particular ecoregion, but different batches can focus on different ecoregions with distinct landscape characteristics. For such a scenario, we study the following questions: (1) How well do DL models trained in homogeneous regions perform when they are transferred to different ecoregions? (2) Does increasing the spatial coverage in the data improve model performance in a given ecoregion? and (3) Can a landslide pixel labeling model be incrementally updated with new data, but without access to the old data and without losing performance on the old data? We address these questions by developing a mechanism for incremental training of semantic segmentation models. We call the resulting extension task-specific model updates (TSMU). A national compilation of landslide inventories by the U.S. Geological Survey (USGS) was used to develop a global database for this study. Our results indicate that the TSMU framework can be used to aid in the creation of new landslide inventories or expanding existing datasets, and also to rapidly develop hazard maps for situational awareness following a widespread landslide event.https://ieeexplore.ieee.org/document/9780028/Catastrophic forgettingcontinual learningdeep learning (DL)domain adaptationecoregionslandslides |
spellingShingle | Savinay Nagendra Daniel Kifer Benjamin Mirus Te Pei Kathryn Lawson Srikanth Banagere Manjunatha Weixin Li Hien Nguyen Tong Qiu Sarah Tran Chaopeng Shen Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Catastrophic forgetting continual learning deep learning (DL) domain adaptation ecoregions landslides |
title | Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates |
title_full | Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates |
title_fullStr | Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates |
title_full_unstemmed | Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates |
title_short | Constructing a Large-Scale Landslide Database Across Heterogeneous Environments Using Task-Specific Model Updates |
title_sort | constructing a large scale landslide database across heterogeneous environments using task specific model updates |
topic | Catastrophic forgetting continual learning deep learning (DL) domain adaptation ecoregions landslides |
url | https://ieeexplore.ieee.org/document/9780028/ |
work_keys_str_mv | AT savinaynagendra constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT danielkifer constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT benjaminmirus constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT tepei constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT kathrynlawson constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT srikanthbanageremanjunatha constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT weixinli constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT hiennguyen constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT tongqiu constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT sarahtran constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates AT chaopengshen constructingalargescalelandslidedatabaseacrossheterogeneousenvironmentsusingtaskspecificmodelupdates |