Task Grouping for Multilingual Text Recognition

Jing Huang, Kevin J. Liang, Rama Kovvuri, Tal Hassner

نتاج البحث: فصل من :كتاب / تقرير / مؤتمرمنشور من مؤتمرمراجعة النظراء

ملخص

Most existing OCR methods focus on alphanumeric characters due to the popularity of English and numbers, as well as their corresponding datasets. On extending the characters to more languages, recent methods have shown that training different scripts with different recognition heads can greatly improve the end-to-end recognition accuracy compared to combining characters from all languages in the same recognition head. However, we postulate that similarities between some languages could allow sharing of model parameters and benefit from joint training. Determining language groupings, however, is not immediately obvious. To this end, we propose an automatic method for multilingual text recognition with a task grouping and assignment module using Gumbel-Softmax, introducing a task grouping loss and weighted recognition loss to allow for simultaneous training of the models and grouping modules. Experiments on MLT19 lend evidence to our hypothesis that there is a middle ground between combining every task together and separating every task that achieves a better configuration of task grouping/separation.

اللغة الأصليةالإنجليزيّة
عنوان منشور المضيفComputer Vision – ECCV 2022 Workshops, Proceedings
المحررونLeonid Karlinsky, Tomer Michaeli, Ko Nishino
ناشرSpringer Science and Business Media Deutschland GmbH
الصفحات297-313
عدد الصفحات17
رقم المعيار الدولي للكتب (المطبوع)9783031250682
المعرِّفات الرقمية للأشياء
حالة النشرنُشِر - 2023
الحدث17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, إسرائيل
المدة: ٢٣ أكتوبر ٢٠٢٢٢٧ أكتوبر ٢٠٢٢

سلسلة المنشورات

الاسمLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
مستوى الصوت13804 LNCS
رقم المعيار الدولي للدوريات (المطبوع)0302-9743
رقم المعيار الدولي للدوريات (الإلكتروني)1611-3349

!!Conference

!!Conference17th European Conference on Computer Vision, ECCV 2022
الدولة/الإقليمإسرائيل
المدينةTel Aviv
المدة٢٣/١٠/٢٢٢٧/١٠/٢٢

ملاحظة ببليوغرافية

Publisher Copyright:
© 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

بصمة

أدرس بدقة موضوعات البحث “Task Grouping for Multilingual Text Recognition'. فهما يشكلان معًا بصمة فريدة.

قم بذكر هذا