From: Ancient mural classification methods based on a multichannel separable network
Parameter | Values | |
---|---|---|
Category | labels_num | 8 |
Batch size | batch_size | 16 |
Learning rate | base_lr | 0.01 |
Patch | max_step | 20,000 |
Save time | snapshot | 2000 |
Hidden unit probability | keep_pro | 0.5 |
Smoothing parameter | ε | 8 |
Train log | train_log_step | 100 |
Test log | Val_log_step | 200 |
Scaling factor | β | 0.2 |