Papers by Siriporn Supratid
The Korean Journal of Mathematics, 2009
We would like to propose Dirichlet-Jordan theorem on the space of summable in measure(). Surely, ... more We would like to propose Dirichlet-Jordan theorem on the space of summable in measure(). Surely, this is a kind of extension of bounded variation([1, 4]), and considered as an application of fuzzy set such that -cut is 0.
This paper presents a comparison study on using softmax, linear discriminant analysis (LDA) and q... more This paper presents a comparison study on using softmax, linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) for object recognition. The least effort is needed for hyper-parameter tuning or selection for all such three classifiers. Convolutional neural network (CNN), using feed-forward-architecture deep learning neural network is employed here for efficient feature extraction and reduction. Then, the extracted, reduced features are fed into the classification comparison. The experiments rely on a small-image CIFAR-10 dataset such that a simple, four convolutional-layer CNN architecture can possibly handle effective feature extraction with hardly over-fitting. Recognition performance evaluations rely on averages of precision, recall, F1 scores and accuracy rates, based on 10-fold cross validation for bias reduction purpose. Such performance measures are implemented under balanced as well as unbalanced --class data, respectively referred to equal and uniform-random-sampling unequal --size class dataset. The results indicate a few bits of recognition performance differences regarding F1 scores as well as accuracy rates among the CNN-LDA, CNN-QDA and CNN-softmax, where the balanced-class and unbalanced-class are separately determined. However, the lowest and the highest of the largest wrong prediction cases are generated by CNN-QDA and CNN-softmax respectively for both balanced and unbalanced-class data.
2023 International Electrical Engineering Congress (iEECON)
2023 International Electrical Engineering Congress (iEECON)
2016 7th International Conference on Intelligent Systems, Modelling and Simulation (ISMS), 2016
Object recognition is one of the research areas which has always attracted the attention of the r... more Object recognition is one of the research areas which has always attracted the attention of the researchers and research community because of its varied application in automation, biometrics, medical diagnosis, surveillance and security systems, defence, Content-based Image Retrieval (CBIR), robotics and intelligent vehicle systems. Though a vigorous research is going on in this field but issues like scale, rotation, illumination invariance, and occlusion, pose and position estimation of objects still draw the attention of researchers. In this paper we have tried to give an overview of the contemporary state of art techniques mainly Featurebased approaches along with the most recent and effective techniques been applied in this area. We have implemented SIFT on COIL dataset and have tried to give a comparative analysis of these techniques.
2022 International Electrical Engineering Congress (iEECON)
This paper focuses on evaluating impacts of large -, medium - and small -size deep residual-learn... more This paper focuses on evaluating impacts of large -, medium - and small -size deep residual-learning convolutional neural network (DRL-CNN): ResNet50, ResNet35 and ResNet17 models on classifying Oxford-102 flowers image dataset with distinct number of 10, 50 and 102 flower classes. The Flowers image classification assessments rely on precision, recall, F1 scores and accuracy rates, averaged over 10-fold cross validation to ensure unbiased experimented results. Confusion matrix is also considered for more detail of results examination. The comparison results indicate the ResNet35 yields 0.201% and 0.706% few better recognition accuracy consecutively over ResNet50 and 17 according to 10-class dataset. For 50-class, 0.060% and 0.211% bits higher accuracy of ResNet35 than ResNet50 and ResNet17 are respectively generated. Whereas, 0.040% and 0.070% a few bits better performance of ResNet35 than ResNet50 and ResNet17 are sequentially attained on 100-class one. Decreasing rate regarding superiority of ResNet35 over ResNet50 and ResNet17 is indicated when increasing class size. However, less than 0.71% higher classification performance is indicated for ResNet35 than ResNet17 for all cases within the scope of this work. Thus, ResNet17 may be preferred to ResNet35 due to approximate 33% higher amount of parameters used in ResNet35 than ResNet17.
The traditional Locally Linear Embedding (LLE) technique was applied for face hallucination. This... more The traditional Locally Linear Embedding (LLE) technique was applied for face hallucination. This technique determines the optimal weights by the fixed number of neighbors for every point. Our previous work, named an adaptive locally linear embedding (ALLE), referred to a modified version of LLE was proposed to apply with frontal view face hallucination; it uses a threshold of similarity for selecting the neighbors of each point. However, frontal face is barely captured in the real world. Therefore, this paper proposes a novel ALLE for multiview face hallucination. The main objective is to generate high quality of frontal and non-frontal face images. The processing steps, according to the proposed method are operated as follows; first, a low resolution (LR) face in one of front, up, down, left or right views is fed as an input; then, the other views of such an LR image are generated by ALLE, which applies a threshold of similarity for selecting the neighbors of each point; and high ...
Fuzzy C-Means (FCM) algorithm is one of the well-known unsupervised clustering techniques. Such a... more Fuzzy C-Means (FCM) algorithm is one of the well-known unsupervised clustering techniques. Such an algorithm can be used for unsupervised image clustering. The different initializations cause different evolutions of the algorithm. Random initializations may lead to improper convergence. This paper proposes FCM algorithm initialized by fixed threshold clustering. The purpose of the algorithm is to retrieve from the database the color JPEG images. Two case studies regard to index or represent the color images by either using color temperature histogram or color histogram vectors. The clustering process produces from such an image index the information, which is a degree of membership for each image. This information would be stored in a database. This paper shows that for both two cases, FCM algorithm initialized by fixed threshold clustering gives more accurate results than FCM with random initialization does.
2022 International Electrical Engineering Congress (iEECON)
This paper investigates effects of shortcut-level amount in ResNet of ResNet (RoR) based upon lig... more This paper investigates effects of shortcut-level amount in ResNet of ResNet (RoR) based upon lightweight or small-size ResNet11 on object recognition using CIFAR-100 image dataset with different sizes of 10, 50, 100 object categories. Recognition performance comparison among the traditional ResNet11, 1–and 2–shortcut-level ResNet of ResNet (1L- and 2L-RoRs) is evaluated, relying on precision, recall, F1 and accuracy scores, averaged over 10-fold cross validation. Such cross validation is performed to ensure unbiased experimental results. Confusion matrix is also considered for more detail of results investigation. The comparison results indicate ResNet11, 1L- and 2L-RoRs provide best recognition accuracy of 96.38%, 98.70% and 98.55% respectively for 10-, 50- and 100–class CIFAR-100 datasets. It is also noticed the high competency of 2L-RoR on maintaining the recognition performance as increasing the number of data classes, relative to those 1L-RoR and ResNet11. In addition, 2L-RoR model employs only 0.780% more parameters than 1L-RoR; whilst 1L-RoR utilizes a few bits 0.392% higher parameters than ResNet11.
In the real world of medical diagnosis, interpretation and integration of the rule-based systems ... more In the real world of medical diagnosis, interpretation and integration of the rule-based systems is significantly necessary. Fuzzy version of ant- classification (FAC) provides a framework of prominent achievement on fuzzy rule-based systems. This is caused by the nature of simplicity, accuracy and comprehensibility belonging to ant-based learning and fuzzy systems as well. However, local optimal traps is still a non-trivial problem during rules generating process. The Particle Swarm Optimization (PSO), a robust stochastic evolutionary algorithm based on the movement and intelligence of swarms indicates outstanding performance on a wide range of applications. This paper proposes PSO-MFAC, which utilizes particle swarm optimization algorithm to find the optimal fuzzy set parameters, associated with the modified fuzzy ant-based classification (MFAC). MFAC is a modified version of the traditional fuzzy ant-based classification in terms of attributes and training cases weighting. The pr...
Journal of Water and Climate Change, 2015
The 2011 monsoon season was exceptionally heavy, leading to extensive and long-lasting flooding i... more The 2011 monsoon season was exceptionally heavy, leading to extensive and long-lasting flooding in the Chao Phraya river basin. Flooding was exacerbated by rapid expansion of urban areas into flood plains and was the costliest natural disaster in the country's history, with direct damages estimated at US$ 45 billion. The present study examines the flood behavior in 2011 and flood impact from changing climate. Two generations of the global climate model (GCM), ensembles CMIP3 and CMIP5, are statistically downscaled through historical 20th century and future projections. The majority of GCMs overestimate the dry spell (in June and July) and underestimate the peak precipitation (in May and September). However, they can simulate the mean precipitation reasonably well. Use of the Multi Model Mean shows continuously increased precipitation from near-future to far-future, while the Multi Model Median shows increased precipitation only for the far-future. These findings in changing prec...
2022 19th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON)
2021 9th International Electrical Engineering Congress (iEECON)
This paper focuses a study on impacts of kernel sizes on different resized image relying on convo... more This paper focuses a study on impacts of kernel sizes on different resized image relying on convolutional neural network (CNN) for object recognition. Two sets of convolutional neural network (CNN) deep learning models: Conv573 and Conv3, based on shallow feed-forward-architecture network are employed here for feature extraction with comparative assessment purpose. The Conv573 refers to CNN with kernel size of 7×7, 5×5 and 3×3; whilst the Conv3 represents that of only 3×3 kernel size; where three and two convolutional layers of 3×3 kernel size are consecutively comparable to one convolutional layer of 7×7 and 5×5 kernel sizes. The experiments rely on different-resized CIFAR-10 image dataset, 50×50, 100×100 and 150×150 pixels for testing with the Conv573 and Conv3 models. For the purpose of bias reduction, recognition performance assessments depend on averages of precision, recall, F1 and accuracy rates, based upon 10-fold cross validation. The results indicate that the greater the size of an image is, the better the recognition accuracy would be based on Conv573, conversely for Conv3. 2.65% and 0.06% recognition accuracy improvement based on Conv573, whereas 0.39% and 1.76% performance decrease based on Conv3 are respectively yielded when resizing image from 50×50 to 100×100 and from 100×100 to 150×150. For 50×50 and 100×100 resized images, Conv3 yields 4.74% and 1.64% better averaged accuracy than Conv573; nevertheless, Conv573 generates 0.18% better averaged accuracy than Conv3 on 150×150 resized ones.
Several clustering researches have attempted to optimize the clustering approaches regarding init... more Several clustering researches have attempted to optimize the clustering approaches regarding initial clusters. The purpose is to alleviate local optima traps. However, such an optimization may possibly not significantly improve the accuracy rate; contrarily it usually generates abundant runtime consumption. In addition, it may cause the emergence of local traps rather than providing the proper clusters initialization. One may turn to focus on the problems of high dimensional, noisy data and outliers hidden in real-world data. Such difficulties can seriously spoil the computation of several types of learning, including clustering. Feature reduction is one of the approaches to relieve such problems. Thereby, this paper proposes a performance comparison using principal component analysis (PCA) and differential evolution (DE) on fuzzy clustering. The purpose relates to evaluating the consequences of feature reduction, compared to those of optimization of the clustering environment. Here, the fuzzy clustering approaches, fuzzy c-means (FCM) and k-harmonic means (KHM) are experimented. FCM and KHM are soft clustering algorithms that retain more information from the original data than those of crisp or hard. PCA, the feature reduction method, is employed as a preprocessing of FCM and KHM for relieving the curse of high-dimensional, noisy data. The performance of the FCM and KHM based on PCA feature extraction, called PCAFCM and PCAKHM are compared with related algorithms, including the FCM and KHM optimized by differential evolution (DE) method. Comparison tests are performed related to 7 well-known benchmark real-world data sets. Within the scope of this study, the superiority of the feature reduction using PCA over DE optimization on FCM and KHM is indicated.
2019 11th International Conference on Information Technology and Electrical Engineering (ICITEE), 2019
Goodman([3]) presented the question of finding a best possible upper bound of the form , where t(... more Goodman([3]) presented the question of finding a best possible upper bound of the form , where t(G) denote the number of triangles in given graph G. In this, the form of squares of degrees is appeared and many researches have been pursued as an application related to this. Here, we would like to deal with corollary related to the results of Nikiforov([6]).
2021 9th International Electrical Engineering Congress (iEECON), 2021
This paper focuses on effects of number and position of auxiliary networks used in inception conv... more This paper focuses on effects of number and position of auxiliary networks used in inception convolutional neural networks (ICNNs) on object recognition. The ICNNs, using six modules of inception- block having two-auxiliary network (2-ICNN), one- auxiliary-network-after-first-inception-block (1F- ICNN), one-auxiliary-network-before-last-inception- block, (1L-ICNN) are experimented here. According to 2- ICNN, two auxiliary networks are inserted after the first inception-block and before the last one; whilst the auxiliary networks before the last inception-block and after the first one are removed according to 1F-ICNN and 1L-ICNN, consecutively. The experiments rely on Oxford-17 and Oxford-102 flower datasets. Recognition performance assessments depend on averages of F1 and accuracy scores, based on 10-fold cross validation for bias reduction purpose. The results indicate that 1F-ICNN yields 81.88% and 86.39% best recognition performance for Oxford-17, containing 17 classes of flower species; whereas, 2-ICNN provides 70.16% and 79.29% best performance for Oxford-102 with 102 classes of species, based on 70 × 70 and 140 × 140 pixels resized images.
Several clustering researches have attempted to optimize the clustering approaches regarding init... more Several clustering researches have attempted to optimize the clustering approaches regarding initial clusters. The purpose is to alleviate local optima traps. However, such an optimization may possibly not significantly improve the accuracy rate; contrarily it usually generates abundant runtime consumption. In addition, it may cause the emergence of local traps rather than providing the proper clusters initialization. One may turn to focus on the problems of high dimensional, noisy data and outliers hidden in real-world data. Such difficulties can seriously spoil the computation of several types of learning, including clustering. Feature reduction is one of the approaches to relieve such problems. Thereby, this paper proposes a performance comparison using principal component analysis (PCA) and differential evolution (DE) on fuzzy clustering. The purpose relates to evaluating the consequences of feature reduction, compared to those of optimization of the clustering environment. Here...
PurposeA major key success factor regarding proficient Bayes threshold denoising refers to noise ... more PurposeA major key success factor regarding proficient Bayes threshold denoising refers to noise variance estimation. This paper focuses on assessing different noise variance estimations in three Bayes threshold models on two different characteristic brain lesions/tumor magnetic resonance imaging (MRIs).Design/methodology/approachHere, three Bayes threshold denoising models based on different noise variance estimations under the stationary wavelet transforms (SWT) domain are mainly assessed, compared to state-of-the-art non-local means (NLMs). Each of those three models, namely D1, GB and DR models, respectively, depends on the most detail wavelet subband at the first resolution level, on the entirely global detail subbands and on the detail subband in each direction/resolution. Explicit and implicit denoising performance are consecutively assessed by threshold denoising and segmentation identification results.FindingsImplicit performance assessment points the first–second best accu...
Uploads
Papers by Siriporn Supratid