Skip to Main content Skip to Navigation
Conference papers

Ré-identification de personne avec perte min-triplet

Abstract : In this paper, we propose a person re-identification method based on a triplet neural network. The person re-identification problem consists to find one identity among a set of people in different places with one or several non-overlapping surveillance cameras. In order to do this, our system extracts first viewpoint invariant features, then uses a triplet neural network to learn a metric. We propose a new loss function for this neural network called min-triplet loss. Finally, we compare our results with the state-of-art on several public person re-identification benchmarks.
Complete list of metadatas

Cited literature [31 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01625496
Contributor : Yiqiang Chen <>
Submitted on : Friday, October 27, 2017 - 4:27:01 PM
Last modification on : Monday, January 20, 2020 - 5:44:07 PM
Document(s) archivé(s) le : Sunday, January 28, 2018 - 4:12:35 PM

File

CHEN_orasis2017.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01625496, version 1

Citation

Yiqiang Chen, Stefan Duffner, Andrei Stoian, Jean-Yves Dufour, Atilla Baskurt. Ré-identification de personne avec perte min-triplet. 16èmes Journées francophones des jeunes chercheurs en vision par ordinateur (ORASIS 2017), Jun 2017, Colleville-sur-Mer, France. ⟨hal-01625496⟩

Share

Metrics

Record views

567

Files downloads

418