Vehicle Re-Identification in Multi-Camera Scenarios Based on Ensembling Deep Learning Features
Abstract
Vehicle re-identification (ReID) across multiple cameras is one of the principal issues in Intelligent Transportation System (ITS). The main challenge that vehicle ReID presents is the large intra-class and small inter-class variability of vehicles appearance, followed by illumination changes, different viewpoints and scales, lack of labelled data and camera resolution. To address these problems, we present a vehicle ReID system that combines different ReID models, including appearance and orientation deep learning features. Additionally, for results refinement re-ranking and a post-processing step taking into account the vehicle trajectory information provided by the CityFlow-ReID dataset are applied.
Cite
Text
Moral et al. "Vehicle Re-Identification in Multi-Camera Scenarios Based on Ensembling Deep Learning Features." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020. doi:10.1109/CVPRW50498.2020.00310Markdown
[Moral et al. "Vehicle Re-Identification in Multi-Camera Scenarios Based on Ensembling Deep Learning Features." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020.](https://mlanthology.org/cvprw/2020/moral2020cvprw-vehicle/) doi:10.1109/CVPRW50498.2020.00310BibTeX
@inproceedings{moral2020cvprw-vehicle,
title = {{Vehicle Re-Identification in Multi-Camera Scenarios Based on Ensembling Deep Learning Features}},
author = {Moral, Paula and García-Martín, Álvaro and Martínez, José M.},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2020},
pages = {2574-2580},
doi = {10.1109/CVPRW50498.2020.00310},
url = {https://mlanthology.org/cvprw/2020/moral2020cvprw-vehicle/}
}