Please use this identifier to cite or link to this item:
https://hdl.handle.net/1959.11/59112
Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Horry, Michael James | en |
dc.contributor.author | Chakraborty, Subrata | en |
dc.contributor.author | Pradhan, Biswajeet | en |
dc.contributor.author | Shulka, Nagesh | en |
dc.contributor.author | Almazroui, Mansour | en |
dc.date.accessioned | 2024-05-08T06:49:33Z | - |
dc.date.available | 2024-05-08T06:49:33Z | - |
dc.date.issued | 2023-06 | - |
dc.identifier.citation | Earth Systems and Environment, v.7, p. 525-540 | en |
dc.identifier.issn | 2509-9434 | en |
dc.identifier.issn | 2509-9426 | en |
dc.identifier.uri | https://hdl.handle.net/1959.11/59112 | - |
dc.description.abstract | <p>High-velocity data streams present a challenge to deep learning-based computer vision models due to the resources needed to retrain for new incremental data. This study presents a novel staggered training approach using an ensemble model comprising the following: (i) a resource-intensive high-accuracy vision transformer; and (ii) a fast training, but less accurate, low parameter-count convolutional neural network. The vision transformer provides a scalable and accurate base model. A convolutional neural network (CNN) quickly incorporates new data into the ensemble model. Incremental data are simulated by dividing the very large So2Sat LCZ42 satellite image dataset into four intervals. The CNN is trained every interval and the vision transformer trained every half interval. We call this combination of a complementary ensemble with staggered training a “two-speed” network. The novelty of this approach is in the use of a staggered training schedule that allows the ensemble model to efciently incorporate new data by retraining the high-speed CNN in advance of the resource-intensive vision transformer, thereby allowing for stable continuous improvement of the ensemble. Additionally, the ensemble models for each data increment out-perform each of the component models, with best accuracy of 65% against a holdout test partition of the RGB version of the So2Sat dataset.</p> | en |
dc.language | en | en |
dc.publisher | Springer Cham | en |
dc.relation.ispartof | Earth Systems and Environment | en |
dc.rights | Attribution 4.0 International | * |
dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | * |
dc.title | Two-Speed Deep-Learning Ensemble for Classifcation of Incremental Land-Cover Satellite Image Patches | en |
dc.type | Journal Article | en |
dc.identifier.doi | 10.1007/s41748-023-00343-3 | en |
dcterms.accessRights | UNE Green | en |
local.contributor.firstname | Michael James | en |
local.contributor.firstname | Subrata | en |
local.contributor.firstname | Biswajeet | en |
local.contributor.firstname | Nagesh | en |
local.contributor.firstname | Mansour | en |
local.profile.school | School of Science and Technology | en |
local.profile.email | schakra3@une.edu.au | en |
local.output.category | C1 | en |
local.record.place | au | en |
local.record.institution | University of New England | en |
local.publisher.place | Switzerland | en |
local.format.startpage | 525 | en |
local.format.endpage | 540 | en |
local.peerreviewed | Yes | en |
local.identifier.volume | 7 | en |
local.access.fulltext | Yes | en |
local.contributor.lastname | Horry | en |
local.contributor.lastname | Chakraborty | en |
local.contributor.lastname | Pradhan | en |
local.contributor.lastname | Shulka | en |
local.contributor.lastname | Almazroui | en |
dc.identifier.staff | une-id:schakra3 | en |
local.profile.orcid | 0000-0002-0102-5424 | en |
local.profile.role | author | en |
local.profile.role | author | en |
local.profile.role | author | en |
local.profile.role | author | en |
local.profile.role | author | en |
local.identifier.unepublicationid | une:1959.11/59112 | en |
dc.identifier.academiclevel | Academic | en |
dc.identifier.academiclevel | Academic | en |
dc.identifier.academiclevel | Academic | en |
dc.identifier.academiclevel | Academic | en |
dc.identifier.academiclevel | Academic | en |
local.title.maintitle | Two-Speed Deep-Learning Ensemble for Classifcation of Incremental Land-Cover Satellite Image Patches | en |
local.relation.fundingsourcenote | Open Access funding enabled and organized by CAUL and its Member Institutions. This research was supported by Defence Australia funding under the AI for Decision Making Initiative project titled “Efective updating of deep learning models with limited new data” (UTS Ref: 210018980). | en |
local.output.categorydescription | C1 Refereed Article in a Scholarly Journal | en |
local.search.author | Horry, Michael James | en |
local.search.author | Chakraborty, Subrata | en |
local.search.author | Pradhan, Biswajeet | en |
local.search.author | Shulka, Nagesh | en |
local.search.author | Almazroui, Mansour | en |
local.open.fileurl | https://rune.une.edu.au/web/retrieve/811c2c37-ba34-484c-afca-9cf2bd60a207 | en |
local.uneassociation | Yes | en |
local.atsiresearch | No | en |
local.sensitive.cultural | No | en |
local.year.published | 2023 | en |
local.fileurl.open | https://rune.une.edu.au/web/retrieve/811c2c37-ba34-484c-afca-9cf2bd60a207 | en |
local.fileurl.openpublished | https://rune.une.edu.au/web/retrieve/811c2c37-ba34-484c-afca-9cf2bd60a207 | en |
local.subject.for2020 | 4601 Applied computing | en |
local.subject.seo2020 | tbd | en |
local.profile.affiliationtype | External Affiliation | en |
local.profile.affiliationtype | UNE Affiliation | en |
local.profile.affiliationtype | External Affiliation | en |
local.profile.affiliationtype | External Affiliation | en |
local.profile.affiliationtype | External Affiliation | en |
local.date.moved | 2024-05-08 | en |
Appears in Collections: | Journal Article School of Science and Technology |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
openpublished/TwoSpeedChakraborty2023JournalArticle.pdf | Published Version | 2.48 MB | Adobe PDF Download Adobe | View/Open |
This item is licensed under a Creative Commons License