A novel data augmentation method for person re-identification

Shuokun Yang; Fucheng You

doi:10.1117/12.3026744

27 March 2024 A novel data augmentation method for person re-identification

Shuokun Yang, Fucheng You

Proceedings Volume 13105, International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2023); 1310546 (2024) https://doi.org/10.1117/12.3026744
Event: 3rd International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2023), 2023, Qingdao, China

Abstract

Existing datasets for person re-identification do not consider the situation where person encounter occlusion and already have bad lighting conditions, in order to solve this problem, this paper conducts an in-depth research on data augmentation methods for person re-identification and proposes a novel data augmentation method that integrates random erasing and local grayscale transformation. Random erasing generates training images with different degree of occlusion, while local grayscale transformation generates training images with deviations caused by lighting changes. Combining the two can effectively reduce the risk of overfitting and improve the generalization ability of the model, and the experimental results show that rank-1 and mAP improve 1.1% and 1.3% compared to the model using only random erasing, and 1.7% and 2.9% compared to the model without data augmentation. Thus the fusion of random erasing and local grayscale transformation is an effective new data augmentation method.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Shuokun Yang and Fucheng You "A novel data augmentation method for person re-identification", Proc. SPIE 13105, International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2023), 1310546 (27 March 2024); https://doi.org/10.1117/12.3026744

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available