Tamil OCR Conversion from Digital Writing Pad Recognition Accuracy Improves through Modified Deep Learning Architectures

Table 5

Performance evaluation of classifiers through statistical measures based on syllable.

References	Classifiers	Accuracy (%)	Precision (%)	Recall (%)	F1 score (%)
[21]	CNN + SALA (character)	86.3	88.4	94.3	91.2
[22]	AlexNet (character)	90.5	92.6	95.5	94.3
[22]	VGG19Net (character)	92.5	93.5	96.7	95.0
[23]	MMU-SNet (syllable)	95.5	98.6	96.7	97.6
Proposed	RTSBA (syllable)	97.4	99.3	98.0	98.7

CNN + SALA, convolutional neural network, self-adaptive lion algorithm; RTSBA, ResNet two-stage bottleneck architecture; MMU-SNet, modified multi-scale segmentation network.