Kuntpong Woraratpanya

Generative Adversarial Networks (GANs) have demonstrated remarkable capability in synthesizing high-quality images from limited data, addressing challenges of data scarcity and diversity in deep learning (DL) training. This is particularly valuable for car damage classification, where real-world datasets are often limited. To mitigate this, we created a custom damaged-undamaged car dataset for training GAN models and generating realistic car damage images. However, evaluating the per-image realism of GAN-generated images remains challenging. Standard GAN metrics, such as Fréchet Inception Distance (FID), Kernel Inception Distance (KID), and Inception Score (IS), provide dataset-level scores but do not assess individual image quality. Meanwhile, Image Quality Assessment (IQA) metrics require reference images, rendering them unsuitable for reference-free scenarios, particularly in unpaired GAN-generated data. To address these limitations—including the practical failure of standard Mahalanobis Distance Learning (MDL) on small or high-dimensional datasets due to non-invertible covariance matrices—we propose Advanced Mahalanobis Distance Learning (AMDL), which incorporates adaptive regularization and pseudo-inverse refinement on deep feature embeddings from pre-trained CNNs. AMDL enables stable and reliable per-image realism assessment under covariance matrix instability, without requiring large datasets or ground-truth references. Our comprehensive evaluation framework involves three procedures: (1) dataset-level evaluation of four GAN models using standard GAN metrics, (2) per-image realism assessment with AMDL, and (3) classifier-based validation with CNN and Vision Transformer (ViT) models (with vs. without AMDL). Experimental results show that AMDL provides precise per-image realism assessment, outperforms existing GAN metrics across datasets, and offers a practical solution for evaluating unpaired GAN-generated images in car damage classification.

Missing data is a critical challenge in industrial data analysis, particularly during anomaly incidents caused by system equipment malfunctions or, more critically, by cyberattacks in industrial systems. It impedes effective imputation and compromises data integrity. Existing statistical and machine learning techniques struggle with heavily missing data, often failing to restore original data characteristics. To address this, we propose Anomaly Signal Imputation Using Latent Coordination Relations, a framework employing a variational autoencoder (VAE) to learn from complete data and establish a robust imputation model based on latent space coordination points. Experimental results from a water treatment testbed show significant improvements in output signal fidelity despite substantial data loss, outperforming conventional techniques.

The deeper the deep learning (DL) models are, the more computational complexity in need of the vast amount of data training requires to get better performance with accurate results. Generative adversarial networks (GANs) have obtained tremendous attention from many researchers with their impressive generation of synthetic instance data via a few source data by alleviating the problems of data scarcity, insufficient data diversity, and producing only limited plausible alternative data using standard data augmentation techniques conforming to the art of low-data-driven DL training in both scratch models, and pre-trained model in a variety of image classification tasks. That is why to defeat the above-referred problems and a lack of publicly available high-quality car-damaged datasets in car damage analysis, we created a custom data with a framework including three different evaluation assessments to generate a synthesized car-damaged dataset as the comparative study of Cycle-Consistent Adversarial Networks (CycleGAN), and Attention-Guided Generative Adversarial Network (AttentionGAN) by transforming one domain to another with our custom car damaged-undamaged dataset. In addition to this, we evaluated our generated car-damaged images based on three different evaluation assessments: firstly using three quantitative GANs metrics such as Inception Score (IS), Fréchet Inception Distance (FID), and Kernel Inception Distance (KID); secondly creating a convolutional neural networks (CNNs) classifier to identify them into real or fake; and finally building a vision-transformer (ViT) classifier to analyze them into damaged or undamaged. After accomplishing our comparative analysis, we can prove that AttentionGAN is better performance than CycleGAN according to all our experimental results.

Speech Emotion Recognition (SER) is becoming a key role in global business today to improve service efficiency, like call center services. Recent SERs were based on a deep learning approach. However, the efficiency of deep learning depends on the number of layers, i.e., the deeper layers, the higher efficiency. On the other hand, the deeper layers are causes of a vanishing gradient problem, a low learning rate, and high time-consuming. Therefore, this paper proposed a redesign of existing local feature learning block (LFLB). The new design is called a deep residual local feature learning block (DeepResLFLB). DeepResLFLB consists of three cascade blocks: LFLB, residual local feature learning block (ResLFLB), and multilayer perceptron (MLP). LFLB is built for learning local correlations along with extracting hierarchical correlations; DeepResLFLB can take advantage of repeatedly learning to explain more detail in deeper layers using residual learning for solving vanishing gradient and reducing overfitting; and MLP is adopted to find the relationship of learning and discover probability for predicted speech emotions and gender types. Based on two available published datasets: EMODB and RAVDESS, the proposed DeepResLFLB can significantly improve performance when evaluated by standard metrics: accuracy, precision, recall, and F1-score.