Table 2.

A comparison to CRNN related works using the IAM Dataset without a pre-trained model. Bold values indicate the best results for the corresponding experimental configurations.

No.	Recognizer Model	CER	WER
1.	[32]	3.72	6.69
2.	[33]	4.80	9.30
3.	[34]	4.70	8.22
4.	[35]	6.34	16.19
		3.50	9.30
5.	[36]	11.08	22.86
		4.88	12.61
6.	[37]	3.30 Parzival	No
7.	[40]	4.27	8.36
8.	[41]	8.5	--
9.	Proposed method	3.40	6.18