Convolutions vs. Sequences: Understanding performances of neural-based methods for automatic Baybayin script recognition

Cerwin Dexter L. Dela Rosa; Kreed Zion Lorenzo G. Lagunilla; Jomari V. Ramos; Austin Kenneth V. San Pedro; Joseph Marvin R. Imperial

doi:10.1117/12.2669907

24 February 2023 Convolutions vs. Sequences: Understanding performances of neural-based methods for automatic Baybayin script recognition

Cerwin Dexter L. Dela Rosa, Kreed Zion Lorenzo G. Lagunilla, Jomari V. Ramos, Austin Kenneth V. San Pedro, Joseph Marvin R. Imperial

Author Affiliations +

Proceedings Volume 12590, Third International Conference on Computer Vision and Information Technology (CVIT 2022); 1259006 (2023) https://doi.org/10.1117/12.2669907
Event: 2022 3rd International Conference on Computer Vision and Information Technology (CVIT 2022), 2022, Beijing, China

Abstract

Common approaches to vision-based tasks such as character and object recognition use Convolutional Neural Networks (CNNs) due to their practicality in processing images and theoretical grounding. In this work, we take a different perspective in the task of Baybayin script recognition by exploring Vision Transformers, a new paradigm for processing images inspired by the Transformer model. We compare performances of CNNs and ViT and analyzed model confidence on a set of test images using Local Interpretable Model-Agnostic Explanations (LIME). Results show that, performance-wise, convolution-based architectures (CNNS) still outperform sequence-based methods (ViT) for discriminating Baybayin scripts with a nearly doubled performance of 84.5% to 48.8% in accuracy respectively.

Citation Download Citation

Cerwin Dexter L. Dela Rosa, Kreed Zion Lorenzo G. Lagunilla, Jomari V. Ramos, Austin Kenneth V. San Pedro, and Joseph Marvin R. Imperial "Convolutions vs. Sequences: Understanding performances of neural-based methods for automatic Baybayin script recognition", Proc. SPIE 12590, Third International Conference on Computer Vision and Information Technology (CVIT 2022), 1259006 (24 February 2023); https://doi.org/10.1117/12.2669907

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