ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization
As satellite imagery collections continue to grow at an astonishing rate, so is the demand for automated and scalable object detection and segmentation. Scaling computational activities demand models that generalize well across various challenges that can hamper progress, including diverse imaging a...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2021-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9565349/ |
| _version_ | 1818942713162104832 |
|---|---|
| author | Dalton Lunga Jacob Arndt Jonathan Gerrand Robert Stewart |
| author_facet | Dalton Lunga Jacob Arndt Jonathan Gerrand Robert Stewart |
| author_sort | Dalton Lunga |
| collection | DOAJ |
| description | As satellite imagery collections continue to grow at an astonishing rate, so is the demand for automated and scalable object detection and segmentation. Scaling computational activities demand models that generalize well across various challenges that can hamper progress, including diverse imaging and geographic conditions, sampling bias in training data, manual ground truth collection, tooling for model reuse and accountability assessment, and poor model training strategies. A great deal of progress has been made on these challenges. We contribute to the improvement through further development of ReSFlow, a workflow that breaks the problem of model generalization into a collection of specialized exploitations. ReSFlow partitions imagery collections into homogeneous buckets equipped with exploitation models trained to perform well under each bucket.s specific context. Essentially, ReSFlow aims for generalization through stratification. Therefore, within a bucket, exploitation is a homogeneous process that mitigates heterogeneity challenges, including the number of training data and data biases that can occur over varied conditions. Furthermore, custom model architectures and rich training strategies effective for within-bucket conditions can be developed. Meanwhile, across buckets, performance metrics support systematic views of the workflow leading to optimal data labeling allocations and indications that further specialization is warranted. Herein, we discuss the formation of models during the framework's “Offline Initialization” stage. Lastly, we exploit the inherent parallelism due to bucketing to introduce model reuse and demonstrate efficacy by reducing an 89-day manual data labeling cost to zero-days in a new area of interest. |
| first_indexed | 2024-12-20T07:15:48Z |
| format | Article |
| id | doaj.art-61619c664a724712bc7852b1bc3f2a2c |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-12-20T07:15:48Z |
| publishDate | 2021-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-61619c664a724712bc7852b1bc3f2a2c2022-12-21T19:48:48ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352021-01-0114104681048310.1109/JSTARS.2021.31190019565349ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model GeneralizationDalton Lunga0https://orcid.org/0000-0003-0054-1141Jacob Arndt1https://orcid.org/0000-0002-1097-0428Jonathan Gerrand2https://orcid.org/0000-0002-2141-7038Robert Stewart3Geospatial Science and Human Security Division, Oak Ridge National Laboratory, Oak Ridge, TN, USAGeospatial Science and Human Security Division, Oak Ridge National Laboratory, Oak Ridge, TN, USAOak Ridge National Laboratory, Oak Ridge, TN, USAGeospatial Science and Human Security Division, Oak Ridge National Laboratory, Oak Ridge, TN, USAAs satellite imagery collections continue to grow at an astonishing rate, so is the demand for automated and scalable object detection and segmentation. Scaling computational activities demand models that generalize well across various challenges that can hamper progress, including diverse imaging and geographic conditions, sampling bias in training data, manual ground truth collection, tooling for model reuse and accountability assessment, and poor model training strategies. A great deal of progress has been made on these challenges. We contribute to the improvement through further development of ReSFlow, a workflow that breaks the problem of model generalization into a collection of specialized exploitations. ReSFlow partitions imagery collections into homogeneous buckets equipped with exploitation models trained to perform well under each bucket.s specific context. Essentially, ReSFlow aims for generalization through stratification. Therefore, within a bucket, exploitation is a homogeneous process that mitigates heterogeneity challenges, including the number of training data and data biases that can occur over varied conditions. Furthermore, custom model architectures and rich training strategies effective for within-bucket conditions can be developed. Meanwhile, across buckets, performance metrics support systematic views of the workflow leading to optimal data labeling allocations and indications that further specialization is warranted. Herein, we discuss the formation of models during the framework's “Offline Initialization” stage. Lastly, we exploit the inherent parallelism due to bucketing to introduce model reuse and demonstrate efficacy by reducing an 89-day manual data labeling cost to zero-days in a new area of interest.https://ieeexplore.ieee.org/document/9565349/Deep-metric learningmeta-learningobject detectionsatellite semantic segmentation |
| spellingShingle | Dalton Lunga Jacob Arndt Jonathan Gerrand Robert Stewart ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Deep-metric learning meta-learning object detection satellite semantic segmentation |
| title | ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization |
| title_full | ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization |
| title_fullStr | ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization |
| title_full_unstemmed | ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization |
| title_short | ReSFlow: A Remote Sensing Imagery Data-Flow for Improved Model Generalization |
| title_sort | resflow a remote sensing imagery data flow for improved model generalization |
| topic | Deep-metric learning meta-learning object detection satellite semantic segmentation |
| url | https://ieeexplore.ieee.org/document/9565349/ |
| work_keys_str_mv | AT daltonlunga resflowaremotesensingimagerydataflowforimprovedmodelgeneralization AT jacobarndt resflowaremotesensingimagerydataflowforimprovedmodelgeneralization AT jonathangerrand resflowaremotesensingimagerydataflowforimprovedmodelgeneralization AT robertstewart resflowaremotesensingimagerydataflowforimprovedmodelgeneralization |