Dietary Supplement Products Registration Verification Application Using Thai FDA Number Images with OpenAI Vision API

Thai Food and Drug Administration (FDA) registration numbers displayed on product labels are essential for verifying the authenticity of dietary supplements. This study aims to develop a Dietary Supplement Verification Application (DSVAPP) that enables consumers to upload images of product labels containing FDA registration numbers to authenticate products. The system uses the OpenAI vision Application Programming Interface (API) to extract registration numbers, thereby reducing error propagation between object detection and Optical Character Recognition (OCR). DSVAPP is evaluated using 118 images and nine combinations of object detection and OCR methods. The results demonstrate that the DSVAPP incorporating the GPT-5.2 vision model achieves superior performance compared to most evaluated methods.