Benchmark Pashto Handwritten Character Dataset and Pashto Object Character Recognition (OCR) Using Deep Neural Network with Rule Activation Function

Table 4

Performance parameters evaluation on training and testing data.

	Parameters	Model 1	Model 2	Model 3
Training	Accuracy	0.993953	0.94348836	0.0358397
	Cross-entropy	0.151699	0.7376458	6.4359164
	Precision	0.506046	0.50597294	0.5006643
	Recall	0.5060465	0.50594584	0.4993175
	F-measure	0.5060465	0.50595939	0.4999900
Testing	Accuracy	0.8767442	0.81604654	0.0302325
	Cross-entropy	3.7494256	4.209053	3.6924555
	Precision	0.505583	0.5052897	0.5013126
	Recall	0.505947	0.50486486	0.4993462
	F-measure	0.5057650	0.50507721	0.5003275