3 Citationer (Scopus)
8 Downloads (Pure)

Resumé

In person re-identification, the purpose is to match persons across, typically, non-overlapping cameras. This introduces challenges such as occlusion and changes in view and light- ing. In order to overcome these challenges, discriminative features are extracted and used in combination with a su- pervised metric learning algorithm. Most often, feature rep- resentations are created from the entire body, causing noisy features if certain parts are occluded. Therefore, we propose a system which applies the same learning algorithm sepa- rately on feature representations from different body parts and late fuses the outputs, to take advantage of situations in which features from certain body parts are more discrim- inative than other. By evaluation on features at three ab- straction levels, we show that the proposed system increase accuracy by up to 19.87% in the case of high-level features. In addition, we also fuse the features at different abstraction levels to further improve results. Experimental results on VIPeR and CUHK03 show similar performance to state-of- the-art with rank-1 accuracies of 52.72% and 61.50%, respec- tively, while results on the datasets PRID450S and CUHK01 show rank-1 accuracies of 78.36% and 73.40%, respectively, improvements of 11.74% and 7.76% compared to state-of- the-art.
OriginalsprogEngelsk
TitelProceedings of the 9th International Conference on Machine Learning and Computing
ForlagAssociation for Computing Machinery
Publikationsdatojun. 2017
Sider385-393
ISBN (Trykt)978-1-4503-4817-1
DOI
StatusUdgivet - jun. 2017
BegivenhedInternational Conference on Machine Learning and Computing - Singapore, Singapore
Varighed: 24 feb. 201726 feb. 2017
Konferencens nummer: 9

Konference

KonferenceInternational Conference on Machine Learning and Computing
Nummer9
LandSingapore
BySingapore
Periode24/02/201726/02/2017
NavnACM International Conference Proceedings Series

Fingerprint

Electric fuses
Learning algorithms
Fusion reactions
Lighting
Cameras

Citer dette

Lejbølle, A. R., Nasrollahi, K., & Moeslund, T. B. (2017). Late Fusion in Part-based Person Re-identification. I Proceedings of the 9th International Conference on Machine Learning and Computing (s. 385-393). Association for Computing Machinery. ACM International Conference Proceedings Series https://doi.org/10.1145/3055635.3056640
Lejbølle, Aske Rasch ; Nasrollahi, Kamal ; Moeslund, Thomas B. / Late Fusion in Part-based Person Re-identification. Proceedings of the 9th International Conference on Machine Learning and Computing . Association for Computing Machinery, 2017. s. 385-393 (ACM International Conference Proceedings Series).
@inproceedings{887ab867efcd495a984a0eb6ca8beba9,
title = "Late Fusion in Part-based Person Re-identification",
abstract = "In person re-identification, the purpose is to match persons across, typically, non-overlapping cameras. This introduces challenges such as occlusion and changes in view and light- ing. In order to overcome these challenges, discriminative features are extracted and used in combination with a su- pervised metric learning algorithm. Most often, feature rep- resentations are created from the entire body, causing noisy features if certain parts are occluded. Therefore, we propose a system which applies the same learning algorithm sepa- rately on feature representations from different body parts and late fuses the outputs, to take advantage of situations in which features from certain body parts are more discrim- inative than other. By evaluation on features at three ab- straction levels, we show that the proposed system increase accuracy by up to 19.87{\%} in the case of high-level features. In addition, we also fuse the features at different abstraction levels to further improve results. Experimental results on VIPeR and CUHK03 show similar performance to state-of- the-art with rank-1 accuracies of 52.72{\%} and 61.50{\%}, respec- tively, while results on the datasets PRID450S and CUHK01 show rank-1 accuracies of 78.36{\%} and 73.40{\%}, respectively, improvements of 11.74{\%} and 7.76{\%} compared to state-of- the-art.",
keywords = "re-identification, late fusion, Neural Network",
author = "Lejb{\o}lle, {Aske Rasch} and Kamal Nasrollahi and Moeslund, {Thomas B.}",
year = "2017",
month = "6",
doi = "10.1145/3055635.3056640",
language = "English",
isbn = "978-1-4503-4817-1",
series = "ACM International Conference Proceedings Series",
publisher = "Association for Computing Machinery",
pages = "385--393",
booktitle = "Proceedings of the 9th International Conference on Machine Learning and Computing",
address = "United States",

}

Lejbølle, AR, Nasrollahi, K & Moeslund, TB 2017, Late Fusion in Part-based Person Re-identification. i Proceedings of the 9th International Conference on Machine Learning and Computing . Association for Computing Machinery, ACM International Conference Proceedings Series, s. 385-393, Singapore, Singapore, 24/02/2017. https://doi.org/10.1145/3055635.3056640

Late Fusion in Part-based Person Re-identification. / Lejbølle, Aske Rasch; Nasrollahi, Kamal; Moeslund, Thomas B.

Proceedings of the 9th International Conference on Machine Learning and Computing . Association for Computing Machinery, 2017. s. 385-393 (ACM International Conference Proceedings Series).

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

TY - GEN

T1 - Late Fusion in Part-based Person Re-identification

AU - Lejbølle, Aske Rasch

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2017/6

Y1 - 2017/6

N2 - In person re-identification, the purpose is to match persons across, typically, non-overlapping cameras. This introduces challenges such as occlusion and changes in view and light- ing. In order to overcome these challenges, discriminative features are extracted and used in combination with a su- pervised metric learning algorithm. Most often, feature rep- resentations are created from the entire body, causing noisy features if certain parts are occluded. Therefore, we propose a system which applies the same learning algorithm sepa- rately on feature representations from different body parts and late fuses the outputs, to take advantage of situations in which features from certain body parts are more discrim- inative than other. By evaluation on features at three ab- straction levels, we show that the proposed system increase accuracy by up to 19.87% in the case of high-level features. In addition, we also fuse the features at different abstraction levels to further improve results. Experimental results on VIPeR and CUHK03 show similar performance to state-of- the-art with rank-1 accuracies of 52.72% and 61.50%, respec- tively, while results on the datasets PRID450S and CUHK01 show rank-1 accuracies of 78.36% and 73.40%, respectively, improvements of 11.74% and 7.76% compared to state-of- the-art.

AB - In person re-identification, the purpose is to match persons across, typically, non-overlapping cameras. This introduces challenges such as occlusion and changes in view and light- ing. In order to overcome these challenges, discriminative features are extracted and used in combination with a su- pervised metric learning algorithm. Most often, feature rep- resentations are created from the entire body, causing noisy features if certain parts are occluded. Therefore, we propose a system which applies the same learning algorithm sepa- rately on feature representations from different body parts and late fuses the outputs, to take advantage of situations in which features from certain body parts are more discrim- inative than other. By evaluation on features at three ab- straction levels, we show that the proposed system increase accuracy by up to 19.87% in the case of high-level features. In addition, we also fuse the features at different abstraction levels to further improve results. Experimental results on VIPeR and CUHK03 show similar performance to state-of- the-art with rank-1 accuracies of 52.72% and 61.50%, respec- tively, while results on the datasets PRID450S and CUHK01 show rank-1 accuracies of 78.36% and 73.40%, respectively, improvements of 11.74% and 7.76% compared to state-of- the-art.

KW - re-identification

KW - late fusion

KW - Neural Network

U2 - 10.1145/3055635.3056640

DO - 10.1145/3055635.3056640

M3 - Article in proceeding

SN - 978-1-4503-4817-1

T3 - ACM International Conference Proceedings Series

SP - 385

EP - 393

BT - Proceedings of the 9th International Conference on Machine Learning and Computing

PB - Association for Computing Machinery

ER -

Lejbølle AR, Nasrollahi K, Moeslund TB. Late Fusion in Part-based Person Re-identification. I Proceedings of the 9th International Conference on Machine Learning and Computing . Association for Computing Machinery. 2017. s. 385-393. (ACM International Conference Proceedings Series). https://doi.org/10.1145/3055635.3056640