KMITL

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

Browse

Search Results

Now showing 1 - 10 of 21

Some of the metrics are blocked by your
consent settings
DDoS Detection Using a Hybrid CNN–RNN Model Enhanced with Multi-Head Attention for Cloud Infrastructure
(2025-11-01)
Sathaporn, Posathip
;
Krungseanmuang, Woranidtha
;
Chaowalittawin, Vasutorn
;
Benjangkaprasert, Chawalit
;
Purahong, Boonchana
Cloud infrastructure supports modern services across different sectors, such as business, education, lifestyle, government and so on. With the high demand for cloud computing, the security of network communication is also an important consideration. Distributed denial-of-service (DDoS) attacks pose a significant threat. Therefore, detection and mitigation are critically important for reliable operation of cloud-based systems. Intrusion detection systems (IDS) play a vital role in detecting and preventing attacks to avoid damage to reliability. This article presents DDoS detection using a convolutional neural network (CNN) and recurrent neural network (RNN) model enhancement with a multi-head attention mechanism for cloud infrastructure protection enhances the contextual relevance and accuracy of the DDoS detection. Preprocessing techniques were applied to optimize model performance, such as information gained to identify important features, normalization, and synthetic minority oversampling technique (SMOTE) to address class imbalance issues. The results were evaluated using confusion metrics. Based on the performance indicators, our proposed method achieves an accuracy of 97.78%, precision of 98.66%, recall of 94.53%, and F1-score of 96.49%. The hybrid model with multi-head attention achieved the best results among the other deep learning models. The model parameter size was moderately lightweight at 413,057 parameters with an inference time in a cloud environment of less than 6 milliseconds, making it suitable for application to cloud infrastructure.
Some of the metrics are blocked by your
consent settings
Duck Egg Crack Detection Using an Adaptive CNN Ensemble with Multi-Light Channels and Image Processing
(2025-07-01)
Chaowalittawin, Vasutorn
;
Krungseanmuang, Woranidtha
;
Sathaporn, Posathip
;
Purahong, Boonchana
Duck egg quality classification is critical in farms, hatcheries, and salted egg processing plants, where cracked eggs must be identified before further processing or distribution. However, duck eggs present a unique challenge due to their white eggshells, which make cracks difficult to detect visually. In current practice, human inspectors use standard white light for crack detection, and many researchers have focused primarily on improving detection algorithms without addressing lighting limitations. Therefore, this paper presents duck egg crack detection using an adaptive convolutional neural network (CNN) model ensemble with multi-light channels. We began by developing a portable crack detection system capable of controlling various light sources to determine the optimal lighting conditions for crack visibility. A total of 23,904 images were collected and evenly distributed across four lighting channels (red, green, blue, and white), with 1494 images per channel. The dataset was then split into 836 images for training, 209 images for validation, and 449 images for testing per lighting condition. To enhance image quality prior to model training, several image pre-processing techniques were applied, including normalization, histogram equalization (HE), and contrast-limited adaptive histogram equalization (CLAHE). The Adaptive MobileNetV2 was employed to evaluate the performance of crack detection under different lighting and pre-processing conditions. The results indicated that, under red lighting, the model achieved 100.00% accuracy, precision, recall, and F1-score across almost all pre-processing methods. Under green lighting, the highest accuracy of 99.80% was achieved using the image normalization method. For blue lighting, the model reached 100.00% accuracy with the HE method. Under white lighting, the highest accuracy of 99.83% was achieved using both the original and HE methods.
Some of the metrics are blocked by your
consent settings
Banana quality classification using lightweight CNN model with microservice integration system
(2025-06-10)
Chaowalittawin, Vasutorn
;
Krungseanmuang, Woranidtha
;
Sathaporn, Posathip
;
Morita, Fuka
;
Archevapanich, Tuanjai
Banana sorting has been performed manually, which often leads to human error due to the high volume and diverse characteristics involved. This paper presents a banana quality classification system using ConsolutechMobileNetV2 (CST-MobileNetV2) to classify banana ripeness into four categories unripe, ripe, overripe, and rotten. A lightweight deep learning model is proposed and integrated with a uniquely designed microservice system to optimize performance while minimizing computational demands. A publicly available dataset containing 13,478 images was used, and the data split into 56% for training, 14% for validation, and 30% for testing. Image normalization and augmentation techniques were applied to enhance the model's robustness. The model's performance was evaluated using a confusion matrix, achieving 98% precision, recall, and F1-score. The proposed model was compared with other deep learning models to benchmark its performance and deployed in different operating systems to evaluate its flexibility and capabilities. The LINE platform was employed as the user interface, enabling practical interaction with users. The system also demonstrated an average response time of 9.25 seconds per image, ensuring efficient processing, delivers high accuracy and scalability making it a practical and efficient solution for automated banana quality classification.
Some of the metrics are blocked by your
consent settings
TinyML Speech Classification Embedded Ai Module for Hand Rehabilitation Device
(2025-01-01)
Numsomran, Arkorn
;
Numsomran, Arjin
;
Chaoraingern, Jutarut
Recent advancements in artificial intelligence and machine learning have significantly improved healthcare, especially in the development of assistive technologies for rehabilitation. This paper introduces a method for hand rehabilitation that utilizes the capabilities of tiny machine learning to enhance speech classification in a rehabilitation device. The proposed method employs a CNN model capable of classifying speech signals that are indicative of different hand movement patterns. Patients emit these speech signals during prescribed hand exercises, which are crucial for their rehabilitation process. The main focus of this study is on training and deploying a speech classification system that can work in the resource-limited environment of TinyML platforms. We detail the process of capturing speech data, preprocessing it, and extracting the most features relevant to different hand movements. Our results show that using TinyML to help with hand rehabilitation works. The method we came up with shows how TinyML could change the way rehabilitative devices are controlled, and it also shows us what personalized and easy-to-use rehabilitation tools might look like in the future.
Some of the metrics are blocked by your
consent settings
Improving the Sound Classification Accuracy Using CNN-LSTM and MFCC with Audio Augmentation for Diagnosing Respiratory Disease
(2025-01-01)
Phankokkruad, Manop
;
Wacharawichanant, Sirirat
Audio is vital information data for understanding various situations. A multitude of sound features can be explained by analysis through the audio signals. Numerous classification methods have been developed to study audio classification. This work studies the improvement of audio classification for the diagnosis of respiratory disease through the integration of audio data augmentation and CNN in conjunction with LSTM (CNN-LSTM). Furthermore, this paper focuses on audio data augmentation and feature extraction in the deep learning approach. This study proposed the CNN-LSTM model to diagnose respiratory disease by learning from the different audio datasets. The results reveal that the CNN-LSTM model attained an accuracy of 81.48%, precision of 0.8340, sensitivity of 0.6948, and F1-score of 0.7225. Considering the achieved F1-score, the CNN-LSTM model demonstrates a high level of diagnotic accuracy. Therefore, all evaluation evaluation parameters collectively indicate the robust performance of the proposed disease classification model.
Some of the metrics are blocked by your
consent settings
Mitigating Racial Bias in Skin Lesion Classification With a Novel Deep Learning-Driven Dataset
(2025-01-01)
Joshi, Priyanka
;
Ruangrit, Chanet
;
Konghuayrob, Poom
(1) Background: A critical issue in the application of Machine Learning (ML) in dermatology is the presence of racial bias in training datasets, which can lead to disparities in diagnostic performance across different skin tones. This paper proposes a mitigation strategy for the BCN20000 dataset, addressing its limitations regarding darker skin lesions by augmenting it with a custom dataset. (2) Method: This was achieved by augmenting the BCN20000 dataset with a custom-developed dataset of dark skin lesion images, generated by combining serial style transfer and latent diffusion-based upscaling. The impact of this augmented dataset was assessed on the accuracy of several established Convolutional Neural Network (CNN) architectures, including DenseNet, ConvNeXt, EfficientNet, RegNet and ResNet. (3) Results: Our findings underscore the necessity of diverse datasets for AI-driven dermatology tools with the best performing model achieving 92% accuracy when trained on the augmented dataset containing additional dark-skin images compared to baseline models trained solely on BCN20000. These results align with prior studies emphasizing the critical role of representative training data in mitigating racial bias in medical AI systems.
Some of the metrics are blocked by your
consent settings
2D and 3D LiDAR with CNN Models for Detecting Sediment Accumulation Underground after Disasters
(2025-01-01)
Krungseanmuang, Woranidtha
;
Morita, Fuka
;
Chaowalittawin, Vasutorn
;
Sathaporn, Posathip
;
Kanamori, Chisato
Global climate change impacts all regions and leads to natural disasters such as typhoons, which cause destruction, debris, and flooding. Postdisaster restoration is a very important activity that is mostly done manually and can be time-consuming and challenging, especially in subterranean environments owing to accumulated objects such as pipes, pillars, and mud distributed in confined underground areas. Therefore, in this study, we aim to utilize emerging AI technologies by comparing deep learning algorithms and evaluating four models for 2D object detection and four for 3D point cloud segmentation for detecting sediment accumulation and navigating around obstacles in underground areas after a disaster. Additionally, a custom dataset was developed to simulate underground disaster scenarios. As a result, the You Only Look Once version 11 (YOLOv11) model achieved the highest mean average precision 50 (mAP50: 91.1%) for general detection within the pillar-pipe dataset, whereas the YOLOv12 model performed the best in detecting pipes (mAP50: 87.7%). In the mud dataset, the YOLOv8 segmentation (YOLOv8-seg) model demonstrated superior performance with mAP50 scores of 93.0% (detection) and 86.4% (segmentation). For 3D point cloud segmentation, PointNet achieved the highest accuracy (98.61%), whereas RandLA-Net was optimal for pipe segmentation, achieving an intersection over union score of 37.1%. These findings highlight AI’s potential to accelerate disaster recovery, reduce manual labor, and ensure faster cleanup. Integrating deep learning models into post-typhoon restoration efforts can enable communities to recover more quickly and efficiently after climate change impacts or disaster events.
Some of the metrics are blocked by your
consent settings
Study of CNN-Based Data Detection in Dual-Layer Bit-Patterned Magnetic Recording Systems
(2025-01-01)
Sangthong, Siraphop
;
Sokjabok, Siwakon
;
Khametong, Anawin
;
Warisarn, Chanon
This paper introduces an innovative data detection system that utilizes convolutional neural networks (CNNs) for dual-layered bit-patterned magnetic systems. Using a mutual-information CNN architecture, the proposed system tackles the challenge of decoding overlapping readback signals from upper and lower layers. The sliding window detection schemes are implemented with input lengths of 6 (2×3) and 14 (2×7) bits, processing oversampled readback signals from a dataset of 1,000,000 bits. Simulation results conducted over a signal-to-noise ratio range of 10 to 24 dBs indicate that the CNN model with a larger input window significantly outperforms smaller input models and conventional partial response maximum likelihood detectors in terms of bit error rate. These findings illustrate the effectiveness of CNN-based detection in enhancing classification accuracy under high-noise conditions, paving the way for future ultra-high-density magnetic recording systems.
Some of the metrics are blocked by your
consent settings
Deep Learning-Based Channel Estimation With 1D CNN for OFDM Systems Under High-Speed Railway Environments
(2025-01-01)
Siriwanitpong, Aphitchaya
;
Sanada, Kosuke
;
Hatano, Hiroyuki
;
Mori, Kazuo
;
Boonsrimuang, Pisit
In OFDM wireless communications, channel estimation performance is compromised in high-speed railway environments owing to extremely fast multipath fading and severe Doppler effect. Recently, a deep learning approach has been employed to improve the channel estimation performance, however it encounters significant challenges due to its high computational complexity. In order to deal with these challenges, this paper proposes channel estimation employing deep learning with one-dimensional convolutional neural network (1D CNN) schemes to enhance conventional least squares (LS) estimation. The first scheme provides better performance compared to conventional LS estimation. However, it is only suitable for OFDM systems with full pilot symbols, leading to decreased transmission efficiency and high complexity. In order to address those problems, the second scheme develops 1D CNN-based channel estimation employing scattered pilot symbols to enhance transmission efficiency and reduce computational complexity. In comparison to conventional LS estimation and deep learning-based channel estimation with bi-gated recurrent unit (bi-GRU), the performance evaluation demonstrates that the proposed 1D CNN-based schemes simultaneously improve channel estimation performance, transmission efficiency, and reduce computational complexity.
Some of the metrics are blocked by your
consent settings
Clamp Dot Image Classification Using Neural Network
(2024-01-01)
Srikam, Krittapak
;
Saenthon, Anakkapon
In this paper, we discuss the classification of images captured by a machine camera while assembling components. To crop out specific points of interest, we employ image processing. Additionally, we utilize deep learning techniques, specifically convolutional neural networks, to identify the type of equipment being assembled. This approach allows us to determine and record specific parts within a device. However, the main challenge of this project is to achieve both high accuracy and the shortest possible prediction time.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results