. 2022 Jan 11;2022:9965426. doi: 10.1155/2022/9965426

Table 2.

Summary of Arabic handwritten characters recognition using CNN model.

References	Year	Dataset	Type (size)	Method	Optimization	Accuracy (%)	Loss (%)
El-Sawy et al. [6]	2017	AHCD	Chars (16,800)	CNN	(i) Minibatch	94.93	5.1

Mudhsh et al. [22]	2017	ADBase	Digits (6.600)	CNN (based on VGG net)	(ii) Dropout	99.6	—
Mudhsh et al. [22]	2017	HACDB	Chars (70.000)	CNN (based on VGG net)	(iii) Data augmentation	97.32	—

Boufenar et al. [23]	2017	OIHACDB	Chars (6.600)	CNN (based on Alexnet)	(i) Dropout	100	—
Boufenar et al. [23]	2017	AHCD	Chars (6.600)	CNN (based on Alexnet)	(ii) Minibatch	99.98	—

Younis [19]	2018	AHCD	Chars (8.737)	CNN	—	97.7	—
Younis [19]	2018	AIA9K	Chars (8.737)	CNN	—	94.8	—

Latif et al. [20]	2018	Mix of handwriting of multiple languages	Chars	CNN	—	99.26	0.02

Altwaijry and Turaiki [13]	2020	Hijja	Chars (47,434)	CNN	—	88	—
Altwaijry and Turaiki [13]	2020	AHCD	Chars (47,434)	CNN	—	97	—

Alrobah &Albahl [21]	2021	Hijja	Chars (47,434)	CNN + SVM	—	96.3	—

Mustapha et al. [24]	2021	AHCD		CDCGAN	—	—	—