Worasait Suwannik

Worasait Suwannik

Auto-generated profile, based on publicly available information

Worasait Suwannik

Kasetsart University

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 2024 in IAES International Journal of Artificial Intelligence (IJ-AI)
Verified
Journal Article
Open access
PDF file
CKChatchai Kasemtaweechok
Chatchai Kasemtaweechok
WSWorasait Suwannik
Worasait Suwannik
Imbalanced datasets are characterized by a substantially smaller number of data points in the minority class compared to the majority class. This imbalance often leads to poor predictive performance of classification models when applied in real-world scenarios. There are three main approaches to handle imbalanced data: over-sampling, under-sampling, and hybrid approach. The over-sampling methods duplicate or synthesize data in the minority class. On the other hand, the under-sampling methods remove majority class data. Hybrid methods combine the noise-removing benefits of under-sampling the majority class with the synthetic minority class creation process of over-sampling. In this research, we applied principal component (PC) analysis, which is normally used for dimensionality reduction, to reduce the amount of majority class data. The proposed method was compared with eight state-of-the-art under-sampling methods across three different classification models: support vector machine, random forest, and AdaBoost. In the experiment, conducted on 35 datasets, the proposed method had higher average values for sensitivity, G-mean, the Matthews correlation coefficient (MCC), and receiver operating characteristic curve (ROC curve) compared to the other under-sampling methods.

25 August 2023
Verified
Conference Paper
Closed access
SSSuppasit Srisaeng
Suppasit Srisaeng
PSPongsakorn Sadakorn
Pongsakorn Sadakorn
KPKritchavat Ploddi
Kritchavat Ploddi
DADarin Areechokechai
Darin Areechokechai
WSWorasait Suwannik
Worasait Suwannik
Introduction: Vector-borne diseases (VBDs) account for around 17% of global infectious diseases, with malaria and dengue being the most prevalent. Effective VBD prevention involves controlling mosquito populations by spraying micro-bubble (MB) insecticide. Quality control measures for MB size are essential, but few centers can assess MB size measurement. Machine learning (ML) is a powerful tool for automating and enhancing image analysis tasks, including measuring MB size. In this study, we train and evaluate ML models using an MB image dataset to assess their performance.Methods: Our experimental study used an MB image dataset labeled by specialists, with images resized and randomly assigned to train, validate, and test datasets. We employed various evaluation metrics, such as recall, precision, and mean average precision (mAP). Image augmentation techniques were applied to increase the number of unqualified MBs. Pretrained models from Detectron2 (D2), MMDetection (MM), and YOLO libraries were used for training and evaluation.Results: We identified severe class and scale imbalances in the dataset. Single-stage and two-stage CNN models demonstrated superior performance in MB detection compared to DCN and transformer-based models. D2 outperformed MM, and training Faster-RCNN R-50 on an augmented dataset improved performance for unqualified MB detection. Considering multiple evaluation metrics, YOLO v7 exhibited the highest accuracy and effectiveness in detecting MBs.Conclusion: Our study revealed the effectiveness of CNN models in MB detection and the importance of selecting appropriate libraries and pretrained weights. Image augmentation improved unqualified MB detection while considering multiple evaluation metrics provided a more comprehensive understanding of model performance. Models like YOLO v7 and Faster-RCNN R-50 and X-101 are recommended for practical MB quality assessment tasks, as they demonstrate high precision and recall performance for qualified and unqualified MBs.

12 June 2016
Verified
Book Chapter
Closed access
PDF file
KKKedsararat Khawyuen
Kedsararat Khawyuen
WSWorasait Suwannik
Worasait Suwannik
MMMontri Maleewong
Montri Maleewong

3 August 2022 in arXiv
Verified
Preprint
Open access
PDF file
WSWorasait Suwannik
Worasait Suwannik
CycleGAN can be used to transfer an artistic style to an image. It does not require pairs of source and stylized images to train a model. Taking this advantage, we propose using randomly generated data to train a machine learning model that can transfer traditional art style to a black and white silhouette image. The result is noticeably better than the previous neural style transfer methods. However, there are some areas for improvement, such as removing artifacts and spikes from the transformed image.

6 December 2023
Verified
Conference Paper
Open access
CHChanarin Heng
Chanarin Heng
WSWorasait Suwannik
Worasait Suwannik
Font style transfer is a challenging task in computer vision, aimed at extracting the visual characteristics such as stroke contrast and apply them to the content image. In this article, we focus on utilizing SkelGAN, a modified version of the U-Net architecture, to transfer font styles from English to Khmer characters. For our experiment, we collected a dataset of 15 calligraphy fonts containing both Khmer and English characters. To assess the performance of the generated images, we used the Structural Similarity Index Measure (SSIM). We compare the generated images with the ground truth images, providing a quantitative assessment of their similarity. Our experimental results indicate that SkelGAN successfully performs style transfer from English calligraphy to Khmer calligraphy, achieving an average SSIM score of 0.93. This score signifies a significant resemblance between the generated and target styles.