Loading…
DAC: Detector-Agnostic Spatial Covariances for Deep Local Features
Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropi...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Conference Proceeding |
| Language: | English |
| Subjects: | |
| Online Access: | Request full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| cited_by | |
|---|---|
| cites | |
| container_end_page | 738 |
| container_issue | |
| container_start_page | 728 |
| container_title | |
| container_volume | |
| creator | Tirado-Garin, Javier Warburg, Frederik Civera, Javier |
| description | Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropic estimate that uses the predicted score at a given pixel location, and a full estimate via the local structure tensor of the learned score maps. Both methods are easy to implement and can be applied to any deep feature detector. We show that these covariances are directly related to errors in feature matching, leading to improvements in downstream tasks, including solving the perspective-n-point problem and motion-only bundle adjustment. Code is available at https://github.com/javrtg/DAC. |
| doi_str_mv | 10.1109/3DV62453.2024.00034 |
| format | conference_proceeding |
| fullrecord | <record><control><sourceid>ieee_CHZPO</sourceid><recordid>TN_cdi_ieee_primary_10550632</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10550632</ieee_id><sourcerecordid>10550632</sourcerecordid><originalsourceid>FETCH-LOGICAL-i134t-a70933800377566912046605040b325cf002527ec492773ea3051706542985d83</originalsourceid><addsrcrecordid>eNotjs1KxDAURqMgOIx9Al30BVpvcnPz4662jgoFF_5shxhTqYzTkkTBt7eDrr7FOXwcxs451JyDvcTuRQlJWAsQsgYAlEessNoaJMADssdsJaSmShtjTlmR0seiCSO5Abti113TXpVdyMHnKVbN-35KefTl4-zy6HZlO327OLq9D6kcpriYYS77yS9oE1z-iiGdsZPB7VIo_nfNnjc3T-1d1T_c3rdNX40cZa6cBotolkStSSnLBUilgEDCKwryw1JFQgcvrdAag0MgrkGRFNbQm8E1u_j7HUMI2zmOny7-bDkQgUKBv0IUR1Q</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>DAC: Detector-Agnostic Spatial Covariances for Deep Local Features</title><source>IEEE Xplore All Conference Series</source><creator>Tirado-Garin, Javier ; Warburg, Frederik ; Civera, Javier</creator><creatorcontrib>Tirado-Garin, Javier ; Warburg, Frederik ; Civera, Javier</creatorcontrib><description>Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropic estimate that uses the predicted score at a given pixel location, and a full estimate via the local structure tensor of the learned score maps. Both methods are easy to implement and can be applied to any deep feature detector. We show that these covariances are directly related to errors in feature matching, leading to improvements in downstream tasks, including solving the perspective-n-point problem and motion-only bundle adjustment. Code is available at https://github.com/javrtg/DAC.</description><identifier>EISSN: 2475-7888</identifier><identifier>EISBN: 9798350362459</identifier><identifier>DOI: 10.1109/3DV62453.2024.00034</identifier><identifier>CODEN: IEEPAD</identifier><language>eng</language><publisher>IEEE</publisher><subject>3D Geometry ; Bundle adjustment ; Codes ; Deep Local Features ; Detectors ; Estimation ; Local features ; Tensors ; Uncertainty ; Uncertainty Quantification ; Visualization</subject><ispartof>2024 International Conference on 3D Vision (3DV), 2024, p.728-738</ispartof><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10550632$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>309,310,780,784,789,790,27925,54555,54932</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10550632$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Tirado-Garin, Javier</creatorcontrib><creatorcontrib>Warburg, Frederik</creatorcontrib><creatorcontrib>Civera, Javier</creatorcontrib><title>DAC: Detector-Agnostic Spatial Covariances for Deep Local Features</title><title>2024 International Conference on 3D Vision (3DV)</title><addtitle>3DV</addtitle><description>Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropic estimate that uses the predicted score at a given pixel location, and a full estimate via the local structure tensor of the learned score maps. Both methods are easy to implement and can be applied to any deep feature detector. We show that these covariances are directly related to errors in feature matching, leading to improvements in downstream tasks, including solving the perspective-n-point problem and motion-only bundle adjustment. Code is available at https://github.com/javrtg/DAC.</description><subject>3D Geometry</subject><subject>Bundle adjustment</subject><subject>Codes</subject><subject>Deep Local Features</subject><subject>Detectors</subject><subject>Estimation</subject><subject>Local features</subject><subject>Tensors</subject><subject>Uncertainty</subject><subject>Uncertainty Quantification</subject><subject>Visualization</subject><issn>2475-7888</issn><isbn>9798350362459</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2024</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNotjs1KxDAURqMgOIx9Al30BVpvcnPz4662jgoFF_5shxhTqYzTkkTBt7eDrr7FOXwcxs451JyDvcTuRQlJWAsQsgYAlEessNoaJMADssdsJaSmShtjTlmR0seiCSO5Abti113TXpVdyMHnKVbN-35KefTl4-zy6HZlO327OLq9D6kcpriYYS77yS9oE1z-iiGdsZPB7VIo_nfNnjc3T-1d1T_c3rdNX40cZa6cBotolkStSSnLBUilgEDCKwryw1JFQgcvrdAag0MgrkGRFNbQm8E1u_j7HUMI2zmOny7-bDkQgUKBv0IUR1Q</recordid><startdate>20240318</startdate><enddate>20240318</enddate><creator>Tirado-Garin, Javier</creator><creator>Warburg, Frederik</creator><creator>Civera, Javier</creator><general>IEEE</general><scope>6IE</scope><scope>6IL</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIL</scope></search><sort><creationdate>20240318</creationdate><title>DAC: Detector-Agnostic Spatial Covariances for Deep Local Features</title><author>Tirado-Garin, Javier ; Warburg, Frederik ; Civera, Javier</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-i134t-a70933800377566912046605040b325cf002527ec492773ea3051706542985d83</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2024</creationdate><topic>3D Geometry</topic><topic>Bundle adjustment</topic><topic>Codes</topic><topic>Deep Local Features</topic><topic>Detectors</topic><topic>Estimation</topic><topic>Local features</topic><topic>Tensors</topic><topic>Uncertainty</topic><topic>Uncertainty Quantification</topic><topic>Visualization</topic><toplevel>online_resources</toplevel><creatorcontrib>Tirado-Garin, Javier</creatorcontrib><creatorcontrib>Warburg, Frederik</creatorcontrib><creatorcontrib>Civera, Javier</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan All Online (POP All Online) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE Xplore (Online service)</collection><collection>IEEE Proceedings Order Plans (POP All) 1998-Present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Tirado-Garin, Javier</au><au>Warburg, Frederik</au><au>Civera, Javier</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>DAC: Detector-Agnostic Spatial Covariances for Deep Local Features</atitle><btitle>2024 International Conference on 3D Vision (3DV)</btitle><stitle>3DV</stitle><date>2024-03-18</date><risdate>2024</risdate><spage>728</spage><epage>738</epage><pages>728-738</pages><eissn>2475-7888</eissn><eisbn>9798350362459</eisbn><coden>IEEPAD</coden><abstract>Current deep visual local feature detectors do not model the spatial uncertainty of detected features, producing suboptimal results in downstream applications. In this work, we propose two post-hoc covariance estimates that can be plugged into any pretrained deep feature detector: a simple, isotropic estimate that uses the predicted score at a given pixel location, and a full estimate via the local structure tensor of the learned score maps. Both methods are easy to implement and can be applied to any deep feature detector. We show that these covariances are directly related to errors in feature matching, leading to improvements in downstream tasks, including solving the perspective-n-point problem and motion-only bundle adjustment. Code is available at https://github.com/javrtg/DAC.</abstract><pub>IEEE</pub><doi>10.1109/3DV62453.2024.00034</doi><tpages>11</tpages></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | EISSN: 2475-7888 |
| ispartof | 2024 International Conference on 3D Vision (3DV), 2024, p.728-738 |
| issn | 2475-7888 |
| language | eng |
| recordid | cdi_ieee_primary_10550632 |
| source | IEEE Xplore All Conference Series |
| subjects | 3D Geometry Bundle adjustment Codes Deep Local Features Detectors Estimation Local features Tensors Uncertainty Uncertainty Quantification Visualization |
| title | DAC: Detector-Agnostic Spatial Covariances for Deep Local Features |
| url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-01T11%3A21%3A32IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_CHZPO&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=DAC:%20Detector-Agnostic%20Spatial%20Covariances%20for%20Deep%20Local%20Features&rft.btitle=2024%20International%20Conference%20on%203D%20Vision%20(3DV)&rft.au=Tirado-Garin,%20Javier&rft.date=2024-03-18&rft.spage=728&rft.epage=738&rft.pages=728-738&rft.eissn=2475-7888&rft.coden=IEEPAD&rft_id=info:doi/10.1109/3DV62453.2024.00034&rft.eisbn=9798350362459&rft_dat=%3Cieee_CHZPO%3E10550632%3C/ieee_CHZPO%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-i134t-a70933800377566912046605040b325cf002527ec492773ea3051706542985d83%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=10550632&rfr_iscdi=true |