รายละเอียดข้อมูลการตีพิมพ์เผยแพร่
ข้อมูลเชิงลึกของบทความวิจัยและผู้แต่ง
ข้อมูลบทความ
| ชื่อบทความ | V.View: A Thai-Language Mobile Application with RT-DETR-Based Banknote Recognition for Users with Visual Impairments |
|---|---|
| ประเภทการตีพิมพ์ | ประชุมวิชาการระดับนานาชาติ |
| ชื่องานประชุมวิชาการ/วารสาร | The 2026 Biomedical Engineering International Conference (BMEiCON-2026) |
| ผู้แต่ง |
Thanate Angsuwatanakul Thapanee Khemanuwong Thanaporn Wongbubpha Puttaraporn Tananon Pattanida S
เอกพล ไพโรจน์ ผู้แต่งหลัก
|
| วันที่ตีพิมพ์/นำเสนอ | 10 ต.ค. 2569 |
| ปีที่ | 2026 |
| ฉบับที่ | 18 |
| หมายเลขหน้า | 1-6 |
| อยู่ในฐานข้อมูล | ประชุมวิชาการระดับนานาชาติ |
| บทคัดย่อ (Abstract) |
Visual impairment can substantially limit independent daily activities, particularly the identification and management of paper currency. In Thailand, visually impaired individuals often rely on family members, caregivers, or the public during financial transactions, potentially reducing privacy, confidence, and autonomy. Although assistive technologies are increasingly available, few are tailored to Thai-language users or designed to recognize Thai banknotes within an integrated mobile platform. This study presents V.View, a Thai-language mobile assistive application that uses artificial intelligence-enabled computer vision to support visually impaired individuals. The application includes obstacle detection, banknote recognition, color identification, face recognition, and text reading. This paper focuses on the Thai Banknote Recognition Module, developed to promote independent financial management. The module applies the Real-Time Detection Transformer (RT-DETR) to detect and classify Thai banknotes captured through a smartphone camera. Recognized denominations are aggregated, and the total value is communicated through real-time Thai voice feedback. RT-DETR was selected for its end-to-end detection capability, absence of post-processing requirements such as Non-Maximum Suppression, and balance between detection accuracy and inference speed. The system was evaluated using five Thai banknote denominations: 20, 50, 100, 500, and 1,000 THB. It achieved 100Percent accuracy in classifying individual banknotes, with per-class precision, recall, and F1-score of 1.000 across all five denominations, and 90% accuracy in calculating the total value of multiple banknotes. A preliminary usability evaluation with a visually impaired participant also indicated that the application was practical and accessible. These findings demonstrate the potential of Thai-language AI-powered assistive technologies to enhance financial autonomy, accessibility, and autonomous living while reducing reliance on external assistance.
|