This study explores the application of deep convolutional neural networks (CNNs) for accurate pill identification, addressing the limitations of traditional human-based methods. Using a dataset of 1,250 images across 10 household remedy drugs, various CNN architectures, including YOLO models, were tested under different conditions. Results showed that natural lighting was optimal for imprinted pills, while a lightbox improved detection for plain pills. The YOLOv5-tiny model demonstrated the best detection accuracy, and efficientNet_b0 achieved the highest classification performance. While the model showed strong results, its generalization is limited by sample size and drug variability. Nonetheless, this approach holds promise for enhancing medication safety and reducing errors in outpatient care.
The increasing complexity of pharmaceutical treatments requires precise pill identification to ensure patient safety. Traditional methods for pill reconciliation rely on human experts, which are time-consuming and prone to errors. Deep Convolutional Neural Networks (CNNs), particularly effective in image processing, offer a promising solution for automating and enhancing these processes.
คณะวิศวกรรมศาสตร์
This cooperative education project aims to enhance speed and facilitate the verification process for stock issuance, transfers, distributions, and receipts in the warehouse. The primary focus is to address issues related to wasted time and delays in operational processes. Through analysis, it was found that SAP, the current system, involves complex processes requiring specialized expertise. Although the company has developed the iWarehouse system to improve efficiency, delays and procedural complexity persist. To resolve these challenges, Power BI was utilized to visualize data related to stock issuance, transfers, distributions, and receipts, allowing warehouse staff to work more efficiently by minimizing waste and accelerating processes. Additionally, Power Automate was integrated to automate the processing of received stock numbers from emails, reducing errors and delays caused by manual data entry. The results of this improvement indicate a significant increase in employee efficiency and a noticeable reduction in wasted time. Upon project completion, the findings and development approach will be provided to the company for further enhancement.
คณะวิศวกรรมศาสตร์
This thesis presents the application of deep learning for object classification. The selected deep learning architectures studied include Convolutional Neural Networks (CNN) and ResNet18. It covers data preparation, feature extraction, parameter tuning for accuracy comparison, and performance evaluation of the selected models. The aim is to propose an efficient model for use in devices that assist visually impaired individuals in classifying indoor objects and providing sound alerts.
คณะสถาปัตยกรรม ศิลปะและการออกแบบ
This study explores the design, production, and installation of 3D-printed modular artificial reefs (3DMARs) at Koh Khai, Chumphon Province, Thailand, through a design thinking framework. Collaborating with SCG Co., Ltd. and the Department of Marine and Coastal Resources, the research establishes design criteria and installation methods, utilizing content analysis and qualitative research. Key principles such as modularity, flexibility, environmental sustainability, and usability are identified. The user-centered approach optimizes the 3DMARs for transport and deployment, enabling local community involvement and fostering sustainable practices. The modular design supports scalability, enhancing marine habitats and coral larval settlement. Furthermore, underwater monitoring techniques enable site-specific data collection, allowing for the generation of digital twin models. This research offers a practical framework for marine ecosystem restoration and empowers coastal communities in Thailand and beyond