A Novel Deep Convolutional Neural Network Based on ResNet-18 and Transfer Learning for Detection of Wood Knot Defects

Table 4

Training parameters.

Related parameter	Value	Meaning
Batch size	128	Number of pictures per training
Learning rate	1e-4	Initial learning rate
Epoch	200	Training iteration times
CUDA	Enable	Computer unified device architecture
CUDNN	Enable	A GPU acceleration library for deep neural networks