Learning local feature descriptors with triplets and shallow convolutional neural networks

Learning local feature descriptors with triplets and shallow convolutional neural networks. Balntas, V., Riba, E., Ponsa, D., & Mikolajczyk, K. British Machine Vision Conference 2016, BMVC 2016, 2016-Septe:119.1-119.11, 2016.

Paper doi abstract bibtex

It has recently been demonstrated that local feature descriptors based on convolutional neural networks (CNN) can significantly improve the matching performance. Previous work on learning such descriptors has focused on exploiting pairs of positive and negative patches to learn discriminative CNN representations. In this work, we propose to utilize triplets of training samples, together with in-triplet mining of hard negatives. We show that our method achieves state of the art results, without the computational overhead typically associated with mining of negatives and with lower complexity of the network architecture. We compare our approach to recently introduced convolutional local feature descriptors, and demonstrate the advantages of the proposed methods in terms of performance and speed. We also examine different loss functions associated with triplets.

@article{
 title = {Learning local feature descriptors with triplets and shallow convolutional neural networks},
 type = {article},
 year = {2016},
 pages = {119.1-119.11},
 volume = {2016-Septe},
 id = {80baa67e-3f34-3ae2-a915-d74ad0746797},
 created = {2022-09-19T10:49:11.130Z},
 file_attached = {true},
 profile_id = {276016a7-2c9d-3507-8888-093db7c54774},
 group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},
 last_modified = {2022-09-19T10:49:44.441Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 folder_uuids = {02fb5526-03ff-44ad-8d5c-42bd496c3100},
 private_publication = {false},
 abstract = {It has recently been demonstrated that local feature descriptors based on convolutional neural networks (CNN) can significantly improve the matching performance. Previous work on learning such descriptors has focused on exploiting pairs of positive and negative patches to learn discriminative CNN representations. In this work, we propose to utilize triplets of training samples, together with in-triplet mining of hard negatives. We show that our method achieves state of the art results, without the computational overhead typically associated with mining of negatives and with lower complexity of the network architecture. We compare our approach to recently introduced convolutional local feature descriptors, and demonstrate the advantages of the proposed methods in terms of performance and speed. We also examine different loss functions associated with triplets.},
 bibtype = {article},
 author = {Balntas, Vassileios and Riba, Edgar and Ponsa, Daniel and Mikolajczyk, Krystian},
 doi = {10.5244/C.30.119},
 journal = {British Machine Vision Conference 2016, BMVC 2016}
}

Downloads: 0

{"_id":"T2oRSSEGotZKa6hg7","bibbaseid":"balntas-riba-ponsa-mikolajczyk-learninglocalfeaturedescriptorswithtripletsandshallowconvolutionalneuralnetworks-2016","author_short":["Balntas, V.","Riba, E.","Ponsa, D.","Mikolajczyk, K."],"bibdata":{"title":"Learning local feature descriptors with triplets and shallow convolutional neural networks","type":"article","year":"2016","pages":"119.1-119.11","volume":"2016-Septe","id":"80baa67e-3f34-3ae2-a915-d74ad0746797","created":"2022-09-19T10:49:11.130Z","file_attached":"true","profile_id":"276016a7-2c9d-3507-8888-093db7c54774","group_id":"5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1","last_modified":"2022-09-19T10:49:44.441Z","read":false,"starred":false,"authored":false,"confirmed":"true","hidden":false,"folder_uuids":"02fb5526-03ff-44ad-8d5c-42bd496c3100","private_publication":false,"abstract":"It has recently been demonstrated that local feature descriptors based on convolutional neural networks (CNN) can significantly improve the matching performance. Previous work on learning such descriptors has focused on exploiting pairs of positive and negative patches to learn discriminative CNN representations. In this work, we propose to utilize triplets of training samples, together with in-triplet mining of hard negatives. We show that our method achieves state of the art results, without the computational overhead typically associated with mining of negatives and with lower complexity of the network architecture. We compare our approach to recently introduced convolutional local feature descriptors, and demonstrate the advantages of the proposed methods in terms of performance and speed. We also examine different loss functions associated with triplets.","bibtype":"article","author":"Balntas, Vassileios and Riba, Edgar and Ponsa, Daniel and Mikolajczyk, Krystian","doi":"10.5244/C.30.119","journal":"British Machine Vision Conference 2016, BMVC 2016","bibtex":"@article{\n title = {Learning local feature descriptors with triplets and shallow convolutional neural networks},\n type = {article},\n year = {2016},\n pages = {119.1-119.11},\n volume = {2016-Septe},\n id = {80baa67e-3f34-3ae2-a915-d74ad0746797},\n created = {2022-09-19T10:49:11.130Z},\n file_attached = {true},\n profile_id = {276016a7-2c9d-3507-8888-093db7c54774},\n group_id = {5ec9cc91-a5d6-3de5-82f3-3ef3d98a89c1},\n last_modified = {2022-09-19T10:49:44.441Z},\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 = {It has recently been demonstrated that local feature descriptors based on convolutional neural networks (CNN) can significantly improve the matching performance. Previous work on learning such descriptors has focused on exploiting pairs of positive and negative patches to learn discriminative CNN representations. In this work, we propose to utilize triplets of training samples, together with in-triplet mining of hard negatives. We show that our method achieves state of the art results, without the computational overhead typically associated with mining of negatives and with lower complexity of the network architecture. We compare our approach to recently introduced convolutional local feature descriptors, and demonstrate the advantages of the proposed methods in terms of performance and speed. We also examine different loss functions associated with triplets.},\n bibtype = {article},\n author = {Balntas, Vassileios and Riba, Edgar and Ponsa, Daniel and Mikolajczyk, Krystian},\n doi = {10.5244/C.30.119},\n journal = {British Machine Vision Conference 2016, BMVC 2016}\n}","author_short":["Balntas, V.","Riba, E.","Ponsa, D.","Mikolajczyk, K."],"urls":{"Paper":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c/file/564b2124-0c34-5c87-433e-38bfcc0df75e/BRP2016.pdf.pdf"},"biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","bibbaseid":"balntas-riba-ponsa-mikolajczyk-learninglocalfeaturedescriptorswithtripletsandshallowconvolutionalneuralnetworks-2016","role":"author","metadata":{"authorlinks":{}},"downloads":0},"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/bfbbf840-4c42-3914-a463-19024f50b30c","dataSources":["2252seNhipfTmjEBQ"],"keywords":[],"search_terms":["learning","local","feature","descriptors","triplets","shallow","convolutional","neural","networks","balntas","riba","ponsa","mikolajczyk"],"title":"Learning local feature descriptors with triplets and shallow convolutional neural networks","year":2016}