Keras 神经网络模型的多输出 loss weight metrics 设置_神经网络多个输出，默认对多个loss函数如何处理

1
2
3

inputs = [input_1 , input_2]
outputs = [classify, segmentation, others]
model = keras.models.Model(inputs, outputs)

1
2
3
4

my_loss = {
    'classify': 'categorical_crossentropy',\
    'segmentation':'binary_crossentropy',\
    'others':my_loss_fun}

1
2
3
4

my_loss_weights = {
    'classify':1,\
    'segmentation':1,\
    'others':10}

1
2
3
4
5

my_metrics ={
    'classify':'acc',\
    'segmentation':[mean_iou,'acc'],\
    'others':['mse','acc']
    }

1

model.compile(optimizer=Adam(lr=config.LEARNING_RATE), loss=my_loss, loss_weights= my_loss_weights, metrics= my_metrics)