TISNet-Enhanced Fully Convolutional Network with Encoder-Decoder Structure for Tongue Image Segmentation in Traditional Chinese Medicine

Table 1

Training parameters of the proposed method.

Hyperparameters	Interpretation
Normalization	Mean centering and standard deviation normalization of the intensities were performed
Cropping	All the images were center-cropped to a pixel size
Optimization	SGD optimizer with the base learning rate 0.01
Learning rate scheduling	The poly learning rate policy was used, where the current learning rate equals to the base one multiplying
Batch size	4
Epoch number size	50
Momentum	0.9
Weight decay	0.0005