KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 11

Some of the metrics are blocked by your
consent settings
Local Sigmoid Method: Non-Iterative Deterministic Learning Algorithm for Automatic Model Construction of Neural Network
(2020-01-01)
Alfarozi, Syukron Abu Ishaq
;
Pasupa, Kitsuchart
;
Sugimoto, Masanori
;
Woraratpanya, Kuntpong
A non-iterative learning algorithm for artificial neural networks is an alternative to optimize the neural network parameters with extremely fast convergence time. Extreme learning machine (ELM) is one of the fastest learning algorithms based on a non-iterative method for a single hidden layer feedforward neural network (SLFN) model. ELM uses a randomization technique that requires a large number of hidden nodes to achieve the high accuracy. This leads to a large and complex model, which is slow at the inference time. Previously, we reported analytical incremental learning (AIL) algorithm, which is a compact model and a non-iterative deterministic learning algorithm, to be used as an alternative. However, AIL cannot grow its set of hidden nodes, due to the node saturation problem. Here, we describe a local sigmoid method (LSM) that is also a sufficiently compact model and a non-iterative deterministic learning algorithm to overcome both the ELM randomization and AIL node saturation problems. The LSM algorithm is based on 'divide and conquer' method that divides the dataset into several subsets which are easier to optimize separately. Each subset can be associated with a local segment represented as a hidden node that preserves local information of the subset. This technique helps us to understand the function of each hidden node of the network built. Moreover, we can use such a technique to explain the function of hidden nodes learned by backpropagation, the iterative algorithm. Based on our experimental results, LSM is more accurate than other non-iterative learning algorithms and one of the most compact models.
Some of the metrics are blocked by your
consent settings
Detecting Text in Manga Using Stroke Width Transform
(2019-04-10)
Piriyothinkul, Boonyarith
;
Pasupa, Kitsuchart
;
Sugimoto, Masanori
The Japanese comic-book style known as manga is becoming a popular topic for researchers. This paper focuses on the problem of detecting text regions in manga pages. Because it is time-consuming and laborious to identify the text regions in images manually, an automatic approach is highly desirable. Here, we propose a new text-detection method for manga using a Stroke Width Transform (SWT) technique in conjunction with a Support Vector Machine (SVM). Conventional SWT-based text-detection techniques perform poorly with manga because both text and non-Text objects have similar characteristics for strokes, lines, and shapes. To better suit manga, we propose modifying the rules for finding letter candidates, which improves the ability to capture text. An SVM is then used to classify image patches into letter and nonletter regions. We compared our proposed framework with a conventional framework and other text-detection methods including deep-learning techniques. In the results, our proposed method achieved the highest F-measure of 0.506.
Some of the metrics are blocked by your
consent settings
Square Wave Quadrature Amplitude Modulation for Visible Light Communication Using Image Sensor
(2019-01-01)
Alfarozi, Syukron Abu Ishaq
;
Pasupa, Kitsuchart
;
Hashizume, Hiromichi
;
Woraratpanya, Kuntpong
;
Sugimoto, Masanori
Most visible light communication (VLC) technologies use a light emitting diode (LED) as a data transmitter and a photodiode as a receiver. In this paper, we alternatively focus on the use of an image sensor or camera as a receiver due to its wide availability. However, the successful use of an image sensor mainly depends on the efficiency of the encoder-decoder and the modulation scheme. Thus, this paper proposes a novel modulation scheme based on a square wave signal called a square wave quadrature amplitude modulation (SW-QAM) method. This method can accommodate different camera settings and overcome the problem of LED flicker that is generally sensed by human eyes when the LED frequency is low. At the transmitter side, multiple LEDs can be used to increase the transmission bit rate, while, at the receiver side, a Wiener filter is used as a complementary technique to SW-QAM for solving the light interference phenomenon due to the closeness of one LED to another. Our experimental results show that the proposed SW-QAM scheme can decode symbols very well either the for close or far communication distances, dark or bright lighting conditions, and single or multiple LED points.
Some of the metrics are blocked by your
consent settings
Robust and Unified VLC Decoding System for Square Wave Quadrature Amplitude Modulation Using Deep Learning Approach
(2019-01-01)
Alfarozi, Syukron Abu Ishaq
;
Pasupa, Kitsuchart
;
Hashizume, Hiromichi
;
Woraratpanya, Kuntpong
;
Sugimoto, Masanori
We have proposed a square wave quadrature amplitude modulation (SW-QAM) scheme for visible light communication (VLC) using an image sensor in our previous work. Here, we propose a robust and unified system by using a neural decoding method. This method offers essential SW-QAM decoding capabilities, such as LED localization, light interference elimination, and unknown parameter estimation, bundled into a single neural network model. This work makes use of a convolutional neural network (CNN) that has a capability in automatic learning of unknown parameters, especially when it deals with images as an input. The neural decoding method can provide good solutions for two difficult conditions that are not covered by our previous SW-QAM scheme: unfixed LED positions and multiple point spread functions (PSFs) of multiple LEDs. Responding to the above solutions, three recent CNN architectures - VGG, ResNet, and DenseNet - are modified to suit our scheme and other two small CNN architectures - VGG-like and MiniDenseNet - are proposed for low computing devices. Our experimental results show that the proposed neural decoding method performs better in terms of error rate than the theoretical decoding, an SW-QAM decoder with a $Wiener$ filter, in different scenarios. Furthermore, we experiment on the problem of moving camera, i.e., the unfixed position of LED points. For this case, a spatial transformer network (STN) layer is added to the neural decoding method for solving the moving camera problem, and the method with the new layer achieves a remarkable result.
Some of the metrics are blocked by your
consent settings
Using image features and eye tracking device to predict human emotions towards abstract images
(2016-01-01)
Pasupa, Kitsuchart
;
Chatkamjuncharoen, Panawee
;
Wuttilertdeshar, Chotiros
;
Sugimoto, Masanori
Nowadays, emotional semantic image retrieval system enables users to access images that they want in a database according to emotional concept. This leads to affective image classification task which recently attracts researchers’ attention. However, different users may experience different emotions depending on where, in the image, they are gazing on. This paper presents an improved prediction method by taking into account the users eye movement as implicit feedback while they are looking at the image. Our experimental results show that using both eye movement information and image feature together to determine users emotion gave more accurate predictions than using image feature alone.
Some of the metrics are blocked by your
consent settings
Hinge loss projection for classification
(2016-01-01)
Alfarozi, Syukron Abu Ishaq
;
Woraratpanya, Kuntpong
;
Pasupa, Kitsuchart
;
Sugimoto, Masanori
Hinge loss is one-sided function which gives optimal solution than that of squared error (SE) loss function in case of classification. It allows data points which have a value greater than 1 and less than −1 for positive and negative classes, respectively. These have zero contribution to hinge function. However, in the most classification tasks, least square (LS) method such as ridge regression uses SE instead of hinge function. In this paper, a simple projection method is used to minimize hinge loss function through LS methods. We modify the ridge regression and its kernel based version i.e. kernel ridge regression so that it can adopt to hinge function instead of using SE in case of classification problem. The results show the effectiveness of hinge loss projection method especially on imbalanced data sets in terms of geometric mean (GM).
Some of the metrics are blocked by your
consent settings
Analytical incremental learning: Fast constructive learning method for neural network
(2016-01-01)
Alfarozi, Syukron Abu Ishaq
;
Setiawan, Noor Akhmad
;
Adji, Teguh Bharata
;
Woraratpanya, Kuntpong
;
Pasupa, Kitsuchart
Extreme learning machine (ELM) is a fast learning algorithm for single hidden layer feed-forward neural network (SLFN) based on random input weights which usually requires large number of hidden nodes. Recently, novel constructive and destructive parsimonious (CP and DP)-ELM which provide the effectiveness generalization and compact hidden nodes have been proposed. However, the performance might be unstable due to the randomization either in ordinary ELM or CP and DP-ELM. In this study, analytical incremental learning (AIL) algorithm is proposed in which all weights of neural network are calculated analytically without any randomization. The hidden nodes of AIL are incrementally generated based on residual error using least square (LS) method. The results show the effectiveness of AIL which has not only smallest number of hidden nodes and more stable but also good generalization than those of ELM, CP and DP-ELM based on seven benchmark data sets evaluation.
Some of the metrics are blocked by your
consent settings
A rapid motion retrieval technique using simple and discrete representation of motion data
(2015-01-01)
Pantuwong, Natapon
;
Takahara, Kensuke
;
Sugimoto, Masanori
In this paper, we propose a rapid motion retrieval technique using dynamic time warping. The frames of the motions are represented by feature vectors whose elements are integer values. The dimensionality of the feature vectors is reduced by using principal component analysis and the values of vector elements are quantized to two bits. A similarity matrix giving distances between the frames is generated for use by dynamic time warping. Preliminary experiments were conducted to find optimum parameter values by evaluating motion retrieval performance. One important feature of the proposed method is that, if the bit length for the frame representation is fixed, the distance between any two frames in any two motions can be found as an element of the similarity matrix without changing its size, which can achieve rapid motion retrieval via dynamic time warping. Experimental comparison with existing methods demonstrated that our proposed technique can complete retrieval tasks over six times faster than a traditional dynamic time warping method, while achieving almost the same levels of accuracy and computation cost as those for the k-d tree method described in [1]. By using simple and discrete representations of frames, the possibilities of achieving rapid retrieval retaining high retrieval accuracy are explored.
Some of the metrics are blocked by your
consent settings
An intuitive computer animation system for enhancing human creativity
(2013-01-01)
Pantuwong, Natapon
;
Sugimoto, Masanori
Computer animation can be used in a wide variety of applications, including storytelling. However, even producing low-quality computer animation requires a significant effort by professional users. Although it should be a tool capable of enhancing creativity in anyone, it is currently too difficult for some users. This paper presents a framework for a computer animation system that can reduce the difficulty of animation authoring, bringing such tasks within the reach of users at all levels. With this framework, users can focus on the story alone, rather than on details of the animation-creation method. The proposed system can therefore enhance users' creativity in story creation. The proposed framework comprises three modules, namely modeling, motion, and animation. Interaction with these modules is via simple actions, enabling computer animation to be created by anyone. A users' study is also presented in this paper to measure the performance of the proposed system in terms of ease of use and the capacity to enhance human creativity. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
A kinetic energy-based feature for unsupervised motion clustering
(2013-01-01)
Nopparit, Suthasinee
;
Pantuwong, Natapon
;
Sugimoto, Masanori
Motion databases usually contain sequences of movements and searching these vast databases is not an easy task. Motion clustering can reduce this difficulty by grouping sample movements into various motion groups containing similar actions. The pose distance is often used as a feature during motion-clustering tasks. However, the main weakness of this strategy is its computational complexity. Query motions are also required to cluster motion sequences. To address these problems, we propose a motion-clustering algorithm based on the use of kinetic energy to cluster sample motions. Our method does not require query motions during the clustering process, so the clustering results can be generated without supervision. Our experimental results confirmed that our proposed method delivered comparable performance to pose distance-based methods, while its computational complexity was significantly lower than that of existing methods. © 2013 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results