Research Article
Tamil OCR Conversion from Digital Writing Pad Recognition Accuracy Improves through Modified Deep Learning Architectures
Table 4
Evaluation of the classifier’s performance through statistical measures based on the dialect.
| Dialect | Accuracy | Precision | Recall | F1 score | | MMU-SNet (%) | RTSBA (%) | MMU-SNet (%) | RTSBA (%) | MMU-SNet (%) | RTSBA (%) | MMU-SNet (%) | RTSBA (%) |
| Madurai Tamil | 83.9 | 85.9 | 96.1 | 99.2 | 86.2 | 86.3 | 90.9 | 92.3 | Tirunelveli Tamil | 87.1 | 89.7 | 96.3 | 98.5 | 89.6 | 90.7 | 92.8 | 94.5 | Coimbatore Tamil | 90.3 | 92.3 | 96.4 | 99.3 | 93.2 | 92.8 | 94.8 | 95.9 | Thanjavur Tamil | 93.5 | 95.5 | 98.6 | 99.3 | 94.7 | 96.1 | 96.6 | 97.6 |
|
|
RTSBA, ResNet two-stage bottleneck architecture; MMU-SNet, modified multi-scale segmentation network.
|