The impacts of training data spatial resolution on deep learning in remote sensing
Deep learning (DL) is ubiquitous in remote sensing analysis with continued evolution in model architectures and advancement of model types. However, DL is still constrained by the need for extensive training datasets, which are costly and time-consuming to produce. One potential solution is adapting...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | Science of Remote Sensing |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2666017224000695 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850218004356268032 |
|---|---|
| author | Christopher Ardohain Songlin Fei |
| author_facet | Christopher Ardohain Songlin Fei |
| author_sort | Christopher Ardohain |
| collection | DOAJ |
| description | Deep learning (DL) is ubiquitous in remote sensing analysis with continued evolution in model architectures and advancement of model types. However, DL is still constrained by the need for extensive training datasets, which are costly and time-consuming to produce. One potential solution is adapting training data annotations from different spatial resolutions, though the feasibility of such an application has yet to be tested. In this study, we explore the effects of using forest boundary training data derived from the 3D Elevation Program (3DEP) at 1.5m resolution and the National Land Cover Database (NLCD) at 30m to compare the effects on DL model performance. Our research covers diverse landscapes across 11 counties in Indiana (∼11,636 km2), developing 36 DL models to assess the impact of spatial resolution, model architectures, land cover, and training chip sizes. Our results show that higher-resolution training data yield more accurate models, regardless of imagery resolution, though the performance gap (F1 score) was limited to ∼2.7% even at its most extreme. We also found significant variation in performance based on land cover, with average F1 scores of 0.923 in homogeneous forested areas compared to 0.684 in complex urban settings. Despite similar training times between data sources, chipping 3DEP data took roughly five times longer. We expect that the findings from this study will assist future research in optimizing the development of DL training datasets, selection of source imagery at the proper resolution given training data availability, and application of appropriate model tuning depending on landscape complexity. |
| format | Article |
| id | doaj-art-005b3115f09a4bfe8051bb932c61ce45 |
| institution | OA Journals |
| issn | 2666-0172 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Science of Remote Sensing |
| spelling | doaj-art-005b3115f09a4bfe8051bb932c61ce452025-08-20T02:07:56ZengElsevierScience of Remote Sensing2666-01722025-06-011110018510.1016/j.srs.2024.100185The impacts of training data spatial resolution on deep learning in remote sensingChristopher Ardohain0Songlin Fei1Corresponding author.; Department of Forestry & Natural Resources, Purdue University, West Lafayette, IN, 47907, USADepartment of Forestry & Natural Resources, Purdue University, West Lafayette, IN, 47907, USADeep learning (DL) is ubiquitous in remote sensing analysis with continued evolution in model architectures and advancement of model types. However, DL is still constrained by the need for extensive training datasets, which are costly and time-consuming to produce. One potential solution is adapting training data annotations from different spatial resolutions, though the feasibility of such an application has yet to be tested. In this study, we explore the effects of using forest boundary training data derived from the 3D Elevation Program (3DEP) at 1.5m resolution and the National Land Cover Database (NLCD) at 30m to compare the effects on DL model performance. Our research covers diverse landscapes across 11 counties in Indiana (∼11,636 km2), developing 36 DL models to assess the impact of spatial resolution, model architectures, land cover, and training chip sizes. Our results show that higher-resolution training data yield more accurate models, regardless of imagery resolution, though the performance gap (F1 score) was limited to ∼2.7% even at its most extreme. We also found significant variation in performance based on land cover, with average F1 scores of 0.923 in homogeneous forested areas compared to 0.684 in complex urban settings. Despite similar training times between data sources, chipping 3DEP data took roughly five times longer. We expect that the findings from this study will assist future research in optimizing the development of DL training datasets, selection of source imagery at the proper resolution given training data availability, and application of appropriate model tuning depending on landscape complexity.http://www.sciencedirect.com/science/article/pii/S2666017224000695Deep learningForest mappingConvolutional neural networkSpatial resolutionTraining data |
| spellingShingle | Christopher Ardohain Songlin Fei The impacts of training data spatial resolution on deep learning in remote sensing Science of Remote Sensing Deep learning Forest mapping Convolutional neural network Spatial resolution Training data |
| title | The impacts of training data spatial resolution on deep learning in remote sensing |
| title_full | The impacts of training data spatial resolution on deep learning in remote sensing |
| title_fullStr | The impacts of training data spatial resolution on deep learning in remote sensing |
| title_full_unstemmed | The impacts of training data spatial resolution on deep learning in remote sensing |
| title_short | The impacts of training data spatial resolution on deep learning in remote sensing |
| title_sort | impacts of training data spatial resolution on deep learning in remote sensing |
| topic | Deep learning Forest mapping Convolutional neural network Spatial resolution Training data |
| url | http://www.sciencedirect.com/science/article/pii/S2666017224000695 |
| work_keys_str_mv | AT christopherardohain theimpactsoftrainingdataspatialresolutionondeeplearninginremotesensing AT songlinfei theimpactsoftrainingdataspatialresolutionondeeplearninginremotesensing AT christopherardohain impactsoftrainingdataspatialresolutionondeeplearninginremotesensing AT songlinfei impactsoftrainingdataspatialresolutionondeeplearninginremotesensing |