What can modern deep-learning models tell us about asphalt road distress classification: an empirical study

Zijie Lin; Shenglin Li

doi:10.1117/12.2671223

29 March 2023 What can modern deep-learning models tell us about asphalt road distress classification: an empirical study

Zijie Lin, Shenglin Li

Author Affiliations +

Proceedings Volume 12594, Second International Conference on Electronic Information Engineering and Computer Communication (EIECC 2022); 125940T (2023) https://doi.org/10.1117/12.2671223
Event: Second International Conference on Electronic Information Engineering and Computer Communication (EIECC 2022), 2022, Xi'an, China

Abstract

Several years have passed, DL methods have made a series of progress. However, within the field of road distress recognition, there are few works using recent methods, especially the transformer. Studying how modern models can help us in this field is important for us to figure out the direction we should move forward to. In this paper, we mainly research the performance of the modern DL model in asphalt road distress classification. GAPs is the dataset we used. We apply the recent SOTA of CNN model and transformer model to find out how far we have come. Totally, the improvement in these years is limited, 7.4% increment of f1 score for the validation set. Some problems like overfitting still need a better solution. And in terms of the dataset, characters like local similarity need to be considered when making a dataset and designing a model.

Citation Download Citation

Zijie Lin and Shenglin Li "What can modern deep-learning models tell us about asphalt road distress classification: an empirical study", Proc. SPIE 12594, Second International Conference on Electronic Information Engineering and Computer Communication (EIECC 2022), 125940T (29 March 2023); https://doi.org/10.1117/12.2671223

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