Repeatability is not enough: Learning affine regions via discriminability. Mishkin, D., Radenović, F., & Matas, J. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11213 LNCS:287-304, 2018.
![Repeatability is not enough: Learning affine regions via discriminability [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper doi abstract bibtex
Paper doi abstract bibtex A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features, that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator – AffNet – trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches. The source codes and trained weights are available at https://github.com/ducha-aiki/affnet.
@article{
title = {Repeatability is not enough: Learning affine regions via discriminability},
type = {article},
year = {2018},
keywords = {Affine shape,Image retrieval,Local features,Loss function},
pages = {287-304},
volume = {11213 LNCS},
id = {7a439934-e56d-3123-87ef-db17d94ff237},
created = {2022-09-19T10:49:11.034Z},
file_attached = {true},
profile_id = {276016a7-2c9d-3507-8888-093db7c54774},
group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},
last_modified = {2022-09-19T10:49:50.050Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
folder_uuids = {02fb5526-03ff-44ad-8d5c-42bd496c3100},
private_publication = {false},
abstract = {A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features, that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator – AffNet – trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches. The source codes and trained weights are available at https://github.com/ducha-aiki/affnet.},
bibtype = {article},
author = {Mishkin, Dmytro and Radenović, Filip and Matas, Jiři},
doi = {10.1007/978-3-030-01240-3_18},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}
}Downloads: 0
{"_id":"9Ymbix7Wjw4Fo4mCC","bibbaseid":"mishkin-radenovi-matas-repeatabilityisnotenoughlearningaffineregionsviadiscriminability-2018","author_short":["Mishkin, D.","Radenović, F.","Matas, J."],"bibdata":{"title":"Repeatability is not enough: Learning affine regions via discriminability","type":"article","year":"2018","keywords":"Affine shape,Image retrieval,Local features,Loss function","pages":"287-304","volume":"11213 LNCS","id":"7a439934-e56d-3123-87ef-db17d94ff237","created":"2022-09-19T10:49:11.034Z","file_attached":"true","profile_id":"276016a7-2c9d-3507-8888-093db7c54774","group_id":"5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1","last_modified":"2022-09-19T10:49:50.050Z","read":false,"starred":false,"authored":false,"confirmed":"true","hidden":false,"folder_uuids":"02fb5526-03ff-44ad-8d5c-42bd496c3100","private_publication":false,"abstract":"A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features, that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator – AffNet – trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches. The source codes and trained weights are available at https://github.com/ducha-aiki/affnet.","bibtype":"article","author":"Mishkin, Dmytro and Radenović, Filip and Matas, Jiři","doi":"10.1007/978-3-030-01240-3_18","journal":"Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)","bibtex":"@article{\n title = {Repeatability is not enough: Learning affine regions via discriminability},\n type = {article},\n year = {2018},\n keywords = {Affine shape,Image retrieval,Local features,Loss function},\n pages = {287-304},\n volume = {11213 LNCS},\n id = {7a439934-e56d-3123-87ef-db17d94ff237},\n created = {2022-09-19T10:49:11.034Z},\n file_attached = {true},\n profile_id = {276016a7-2c9d-3507-8888-093db7c54774},\n group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},\n last_modified = {2022-09-19T10:49:50.050Z},\n read = {false},\n starred = {false},\n authored = {false},\n confirmed = {true},\n hidden = {false},\n folder_uuids = {02fb5526-03ff-44ad-8d5c-42bd496c3100},\n private_publication = {false},\n abstract = {A method for learning local affine-covariant regions is presented. We show that maximizing geometric repeatability does not lead to local regions, a.k.a features, that are reliably matched and this necessitates descriptor-based learning. We explore factors that influence such learning and registration: the loss function, descriptor type, geometric parametrization and the trade-off between matchability and geometric accuracy and propose a novel hard negative-constant loss function for learning of affine regions. The affine shape estimator – AffNet – trained with the hard negative-constant loss outperforms the state-of-the-art in bag-of-words image retrieval and wide baseline stereo. The proposed training process does not require precisely geometrically aligned patches. The source codes and trained weights are available at https://github.com/ducha-aiki/affnet.},\n bibtype = {article},\n author = {Mishkin, Dmytro and Radenović, Filip and Matas, Jiři},\n doi = {10.1007/978-3-030-01240-3_18},\n journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}\n}","author_short":["Mishkin, D.","Radenović, F.","Matas, J."],"urls":{"Paper":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c/file/0fc22d42-4fcf-366d-1420-c00a3fcdf2ae/Dmytro_Mishkin_Repeatability_Is_Not_ECCV_2018_paper.pdf.pdf"},"biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","bibbaseid":"mishkin-radenovi-matas-repeatabilityisnotenoughlearningaffineregionsviadiscriminability-2018","role":"author","keyword":["Affine shape","Image retrieval","Local features","Loss function"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","dataSources":["2252seNhipfTmjEBQ"],"keywords":["affine shape","image retrieval","local features","loss function"],"search_terms":["repeatability","enough","learning","affine","regions","via","discriminability","mishkin","radenović","matas"],"title":"Repeatability is not enough: Learning affine regions via discriminability","year":2018}