Research Article
CAPTCHA Recognition Method Based on CNN with Focal Loss
Table 2
Comparison of recognition accuracy of different CAPTCHA.
| Type | AlexNet (%) | VGG-16 (%) | GoogleNet (%) | ResNet (%) | Our method (%) |
| CNKI CAPTCHA | 96.95 | 96.58 | 97.34 | 98.76 | 99 | ZhengFang CAPTCHA | 96.25 | 95.89 | 95.65 | 96.2 | 98.5 | Random generated CAPTCHA | 94.38 | 95.89 | 96.86 | 97.25 | 97.84 |
|
|