Research Article
U-DAVIS-Deep Learning Based Arm Venous Image Segmentation Technique for Venipuncture
Table 3
Performance of the U-Net model with varying hyper-parameters.
| ā | Epochs | Learning rate | Activation function | Metric | 50 | 100 | 200 | 0.0001 | 0.0005 | 0.001 | Tanh | Sigmoid | ReLU |
| Dice coefficient | 0.52 | 0.685 | 0.678 | 0.685 | 0.572 | 0.542 | 0.532 | 0.468 | 0.685 | IoU | 0.818 | 0.893 | 0.895 | 0.893 | 0.853 | 0.847 | 0.826 | 0.784 | 0.893 | PSNR ratio | 0.619 | 0.751 | 0.741 | 0.751 | 0.596 | 0.583 | 0.581 | 0.505 | 0.751 |
|
|