Publication Details

BoxCars: 3D Boxes as CNN Input for Improved Fine-Grained Vehicle Recognition

SOCHOR Jakub, HEROUT Adam and HAVEL Jiří. BoxCars: 3D Boxes as CNN Input for Improved Fine-Grained Vehicle Recognition. In: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas: IEEE Computer Society, 2016, pp. 3006-3015. ISBN 978-1-4673-8851-1. ISSN 1063-6919. Available from: http://ieeexplore.ieee.org/document/7780697/
Czech title
BoxCars: 3D Boxes jako vstup pro CNN zlepšující fine-grained klasifikaci automobilů
Type
conference paper
Language
english
Authors
Sochor Jakub, Ing. (DCGM FIT BUT)
Herout Adam, prof. Ing., Ph.D. (DCGM FIT BUT)
Havel Jiří, Ing., Ph.D. (DCGM FIT BUT)
URL
Keywords

Fine-grained recognition, vehicles, CNN, input modification

Abstract

We are dealing with the problem of fine-grained vehicle make&model recognition and verification. Our contribution is showing that extracting additional data from the video stream - besides the vehicle image itself - and feeding it into the deep convolutional neural network boosts the recognition performance considerably. This additional information includes: 3D vehicle bounding box used for "unpacking" the vehicle image, its rasterized low-resolution shape, and information about the 3D vehicle orientation. Experiments show that adding such information decreases classification error by 26% (the accuracy is improved from 0.772 to 0.832) and boosts verification average precision by 208% (0.378 to 0.785) compared to baseline pure CNN without any input modifications. Also, the pure baseline CNN outperforms the recent state of the art solution by 0.081. We provide an annotated set "BoxCars" of surveillance vehicle images augmented by various automatically extracted auxiliary information. Our approach and the dataset can considerably improve the performance of traffic surveillance systems.

Published
2016
Pages
3006-3015
Journal
Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, no. 6, ISSN 1063-6919
Proceedings
The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Conference
Computer Vision and Pattern Recognition 2016, Las Vegas, US
ISBN
978-1-4673-8851-1
Publisher
IEEE Computer Society
Place
Las Vegas, US
DOI
UT WoS
000400012303008
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB11103,
   author = "Jakub Sochor and Adam Herout and Ji\v{r}\'{i} Havel",
   title = "BoxCars: 3D Boxes as CNN Input for Improved Fine-Grained Vehicle Recognition",
   pages = "3006--3015",
   booktitle = "The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)",
   journal = "Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition",
   number = 6,
   year = 2016,
   location = "Las Vegas, US",
   publisher = "IEEE Computer Society",
   ISBN = "978-1-4673-8851-1",
   ISSN = "1063-6919",
   doi = "10.1109/CVPR.2016.328",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/11103"
}
Back to top