To identify eligible applicants, universities must validate 300,000 academic transcripts yearly through the Thailand Central Admission System, or TCAS. However, in the initial round of TCAS, verification procedures face several problems, including a week-long completion time, human processing mistakes, and a schedule conflict in data sharing with the Ministry of Education. Speeding up verification diminishes accuracy, forcing human inspection of each document to meet the deadline. Therefore, we created a platform to verify and certify grades and credits according to criteria. We segment images with YOLOv8 to obtain column contents. After testing many generative AI algorithms to extract transcript data, Tesseract OCR was the most accurate. We repeat OCR with data correction till it meets the requirements. Tesseract OCR had the greatest accuracy of 93% for computing average grade and total credit by topic and subject group in the platform assessment findings. It can handle more work and is 85% faster than manual processing.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Extract Senior High School Transcripts by Integrating OCR and Yolo8

  • Suphanat Sroyphet,
  • Lerdphipat Kanjanarungroj,
  • Pinpawat Limsuwat,
  • Somchoke Ruengittinun,
  • Soontharee Koompairojn

摘要

To identify eligible applicants, universities must validate 300,000 academic transcripts yearly through the Thailand Central Admission System, or TCAS. However, in the initial round of TCAS, verification procedures face several problems, including a week-long completion time, human processing mistakes, and a schedule conflict in data sharing with the Ministry of Education. Speeding up verification diminishes accuracy, forcing human inspection of each document to meet the deadline. Therefore, we created a platform to verify and certify grades and credits according to criteria. We segment images with YOLOv8 to obtain column contents. After testing many generative AI algorithms to extract transcript data, Tesseract OCR was the most accurate. We repeat OCR with data correction till it meets the requirements. Tesseract OCR had the greatest accuracy of 93% for computing average grade and total credit by topic and subject group in the platform assessment findings. It can handle more work and is 85% faster than manual processing.