Improved Crack Detection and Recognition Based on Convolutional Neural Network

Figure 4

Printed loss value. Each iteration prints a loss of 100 times and evaluates the model once per 1000 iterations using the validation data set.