Improving person re-identification using KCCA
We provided the implementation of the paper Giuseppe Lisanti , Iacopo Masi , Alberto Del Bimbo, “Matching People across Camera Views using Kernel Canonical Correlation Analysis”, Eighth ACM/IEEE International Conference on Distributed Smart Cameras, 2014.
Requirements
You need the following software to run the code:
- MATLAB (Windows, Unix version is the same)
- Hardoon KCCA code package. (4.3 KB)
- Descriptors computed as described in the paper for the VIPeR and PRID dataset. (229 MB)
Please, note that the code is automatically attempting to download third-party libraries and data.
Download
The code is on Github please download it here
Demo Example
To run our code just run
demo_reid_kcca.m
You can change the dataset and enable CCA comparison with the following parameters:
>Computing Trial 1... >Applying Kernel to Train and Test... >Computing KCCA on the training set... Centering Kx and Ky Decomposing Kernel with PGSO Computing nbeta from nalpha >Projecting the test data... >Computing distances... >Evaluating results... >Computing Trial 2...
Citation
Please cite our paper with the following bibtex if you use our dataset:
@article{lisanti:icdsc14,
author = {Lisanti, Giuseppe and Masi, Iacopo and {Del Bimbo}, Alberto},
title = {Matching People across Camera Views using Kernel Canonical Correlation Analysis},
booktitle = {Eighth ACM/IEEE International Conference on Distributed Smart Cameras},
year = {2014}, }
and Hardoon’s paper:
@article{hardoon:cca,
author = {Hardoon, David and Szedmak, Sandor and {Shawe-Taylor}, John},
title = {Canonical Correlation Analysis: An Overview with Application to Learning Methods},
booktitle = {Neural Computation},
volume = {Volume 16 (12)},
pages = {2639--2664},
year = {2004}, }
References
[1] Giuseppe Lisanti , Iacopo Masi , Alberto Del Bimbo, Matching People across Camera Views using Kernel Canonical Correlation Analysis”, Eighth ACM/IEEE International Conference on Distributed Smart Cameras, 2014.
License
KCCA-ReId code is Copyright (c) 2014 of Giusppe Lisanti and Iacopo Masi <giuseppe.lisanti, iacopo.masi>@unifi.it. Media Integration and Communication Center (MICC), University of Florence.