Research Article
Surface Defect Detection Method Based on Improved Semisupervised Multitask Generative Adversarial Network
Table 2
The testing performance of each model.
| Model | In the dataset expanded by DCGAN | In the dataset expanded by iSSMT-GAN | AlexNet model (%) | VGG-19 model (%) | MobileNet v3 model (%) | RegNet model (%) | AlexNet model (%) | VGG-19 model (%) | MobileNet v3 model (%) | RegNet model (%) |
| Accuracy | 95.31 | 90.62 | 93.01 | 90.62 | 98.43 | 93.10 | 95.31 | 93.75 | F1 score | 95.83 | 84.34 | 97.87 | 91.44 | 98.17 | 93.40 | 94.72 | 92.66 | Recall rate | 95.10 | 83.16 | 98.42 | 91.58 | 97.77 | 93.71 | 95.82 | 92.50 | Precision | 97.08 | 88.06 | 93.01 | 91.58 | 98.75 | 94.04 | 97.72 | 93.98 |
|
|