Order Article Reprints
Journal: Mathematics, 2024
Volume: 12
Number: 2256
Article:
End-to-End Deep Learning Framework for Arabic Handwritten Legal Amount Recognition and Digital Courtesy Conversion
Authors:
by
Hakim A. Abdo, Ahmed Abdu, Mugahed A. Al-Antari, Ramesh R. Manza, Muhammed Talo, Yeong Hyeon Gu and Shobha Bawiskar
Link:
https://www.mdpi.com/2227-7390/12/14/2256
MDPI offers high quality article reprints with convenient shipping to destinations worldwide. Each reprint features a 270 gsm bright white cover
and 105 gsm premium white paper, bound with two stitches for durability and printed in full color. The cover design is customized to your article
and designed to be complimentary to the journal.
![](mailto:?&subject=From%20MDPI%3A%20%22End-to-End%20Deep%20Learning%20Framework%20for%20Arabic%20Handwritten%20Legal%20Amount%20Recognition%20and%20Digital%20Courtesy%20Conversion"&body=https://www.mdpi.com/2877770%3A%0A%0AEnd-to-End%20Deep%20Learning%20Framework%20for%20Arabic%20Handwritten%20Legal%20Amount%20Recognition%20and%20Digital%20Courtesy%20Conversion
Arabic%20handwriting%20recognition%20and%20conversion%20are%20crucial%20for%20financial%20operations%2C%20particularly%20for%20processing%20handwritten%20amounts%20on%20cheques%20and%20financial%20documents.%20Compared%20to%20other%20languages%2C%20research%20in%20this%20area%20is%20relatively%20limited%2C%20especially%20concerning%20Arabic.%20This%20study%20introduces%20an%20innovative%20AI-driven%20method%20for%20simultaneously%20recognizing%20and%20converting%20Arabic%20handwritten%20legal%20amounts%20into%20numerical%20courtesy%20forms.%20The%20framework%20consists%20of%20four%20key%20stages.%20First%2C%20a%20new%20dataset%20of%20Arabic%20legal%20amounts%20in%20handwritten%20form%20%28%E2%80%9C.png%E2%80%9D%20image%20format%29%20is%20collected%20and%20labeled%20by%20natives.%20Second%2C%20a%20YOLO-based%20AI%20detector%20extracts%20individual%20legal%20amount%20words%20from%20the%20entire%20input%20sentence%20images.%20Third%2C%20a%20robust%20hybrid%20classification%20model%20is%20developed%2C%20sequentially%20combining%20ensemble%20Convolutional%20Neural%20Networks%20%28CNNs%29%20with%20a%20Vision%20Transformer%20%28ViT%29%20to%20improve%20the%20prediction%20accuracy%20of%20single%20Arabic%20words.%20Finally%2C%20a%20novel%20conversion%20algorithm%20transforms%20the%20predicted%20Arabic%20legal%20amounts%20into%20digital%20courtesy%20forms.%20The%20framework%E2%80%99s%20performance%20is%20fine-tuned%20and%20assessed%20using%205-fold%20cross-validation%20tests%20on%20the%20proposed%20novel%20dataset%2C%20achieving%20a%20word%20level%20detection[...]){kind=link}