总结一下，我的模型是2d Unet，损失函数是Dice
以纪元350开始，每个过程最多需要3分钟，批量大小= 10
有人可以帮助您确定问题吗？我无法发布损失图，但是它是锯齿形的，并且验证损失范围为（0.61-0.70）
培训细节：
输入=（12000,144,160,4）和9000以培训3000进行验证。
k=keras.initializers.RandomNormal(mean=0.0, stddev=0.01)
inputs = Input((144,160,4))
n_classes=4
c1 = Conv2D(64, (3, 3), activation='elu', kernel_initializer=k, padding='same') (inputs)
c1 = Dropout(0.1) (c1)
c1 = Conv2D(64, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c1)
p1 = MaxPooling2D((2, 2)) (c1)
c2 = Conv2D(128, (3, 3), activation='elu', kernel_initializer=k, padding='same') (p1)
c2 = Dropout(0.1) (c2)
c2 = Conv2D(128, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c2)
p2 = MaxPooling2D((2, 2)) (c2)
c3 = Conv2D(256, (3, 3), activation='elu', kernel_initializer=k, padding='same') (p2)
c3 = Dropout(0.2) (c3)
c3 = Conv2D(256, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c3)
p3 = MaxPooling2D((2, 2)) (c3)
c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer=k, padding='same') (p3)
c4 = Dropout(0.2) (c4)
c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c4)
p4 = MaxPooling2D(pool_size=(2, 2)) (c4)
c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer=k, padding='same') (p4)
c5 = Dropout(0.3) (c5)
c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c5)
u6 = Conv2DTranspose(512, (2, 2), strides=(2, 2), padding='same') (c5)
u6 = concatenate([u6, c4])
c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer=k, padding='same') (u6)
c6 = Dropout(0.2) (c6)
c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c6)
u7 = Conv2DTranspose(256, (2, 2), strides=(2, 2), padding='same') (c6)
u7 = concatenate([u7, c3])
c7 = Conv2D(256, (3, 3), activation='elu', kernel_initializer=k, padding='same') (u7)
c7 = Dropout(0.2) (c7)
c7 = Conv2D(256, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c7)
u8 = Conv2DTranspose(128, (2, 2), strides=(2, 2), padding='same') (c7)
u8 = concatenate([u8, c2])
c8 = Conv2D(128, (3, 3), activation='elu', kernel_initializer=k, padding='same') (u8)
c8 = Dropout(0.1) (c8)
c8 = Conv2D(128, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c8)
u9 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same') (c8)
u9 = concatenate([u9, c1], axis=3)
c9 = Conv2D(64, (3, 3), activation='elu', kernel_initializer=k, padding='same') (u9)
c9 = Dropout(0.1) (c9)
c9 = Conv2D(64, (3, 3), activation='elu', kernel_initializer=k, padding='same') (c9)
outputs = Conv2D(1, (1, 1), activation='sigmoid') (c9)
model = Model(inputs=[inputs],outputs=[outputs])
model.compile(optimizer=Adam(lr=1e-04), loss=dice_coefficient_loss, metrics=[dice_coefficient])
任何有关模型或解释的帮助/建议都将受到赞赏。