Research Article
Enhancing the Power of CNN Using Data Augmentation Techniques for Odia Handwritten Character Recognition
Table 12
Performance comparison of data augmentation techniques.
| Reference | Dataset | Number of classes | Classifier | Accuracy without augmentation | Accuracy with augmentation (%) |
| [26] | BanglaLekha-isolated | 50 | CNN | 91.81% | 95.25 | [21] | ISI image handwritten Oriya numerals | 10 | GAN | 96.40% | 97.30 | [21] | ISI image handwritten Bangla numerals | 10 | GAN | 98.35% | 98.55 | [21] | ISI image handwritten Devanagari numerals | 10 | GAN | 98.48% | 98.56 | Proposed work | NITROHCSv1.0 Odia characters | 47 | CNN | 97.76% | 98.91 | Proposed work | ISI image handwritten Oriya numerals | 10 | CNN | 97.39 | 98.60 |
|
|