|
Towards a practical use of Gaussian of Gaussian descriptor for person re-identification |
| We make use of Gaussian of Gaussian (GOG) in practical by providing simultaneously two improvements.
First, we re-implement and perform intensive experiments to select the optimal values of the parameters using in GOG feature extraction.
Second, we propose and apply pre-processing techniques on person images.
The experimental results show that the proposed approach allows to extract GOG 2 times faster than the available source code and achieve remarkably high accuracy for person ReID.
The obtained accuracies at rank-1 on VIPeR dataset are 51.75% (with background) and 55.79% (without background).
Fig 1. The proposed framework for person re-identification.
Fig 2. Images obtained after preprocessing techniques.
Fig 3. Obtained CMC curves on VIPeR dataset with our new implementation and the authors source code.
Fig 4. Computation time for each step in our new implementation.
Fig 5. Obtained results on VIPeR dataset when applying the preprocessing technique.
Fig 6. Obtained results on VIPeR dataset without background information. |
| Download: - VIPeR images without background - GOG features - Code for extracing GOG in C++, XQDA metric learning, Retinex and Foreground images [Link] Please send email to Thi-Lan.Le@mica dot edu dot vn |
| References: |
| [1] NGUYEN Thuy-Binh, TRAN Duc-Long, LE Thi-Lan, Pham Thi-Thanh Thuy, Doan Huong-Giang, Towards a practical use of Gaussian of Gaussian descriptor for person re-identification, Submitted to the 5th NAFOSTED Conference on Information and Computer Science (NICS 2018) |
| Home page |
