Question

我有一个NN，尝试使用另一个图像输入来预测一个图像输出。输入是具有3个通道（RGB）的图像，输出类似于同一张灰度图像[示例]（输出是图像RGB的深度，更精确地说）。

[在此处输入图片说明] [输入和输出期望示例]

我尝试使用 U-Net [在此处输入图像描述] [U-Net图像]，更改类似数量的神经元。 U-Net用于图像分割，但是我尝试使用其他图像进行预测。

我无法更改批量大小，而且这一数量非常有限（我的计算机不允许更多数量）。只需 BATCH_SIZE = 4 ；
要进行编译，我使用 Adam 而不进行额外配置（默认情况下），并且;
LOSS = 短信

https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/

inputs = Input((IMG_HEIGHT, IMG_WIDTH, RGB_CHANNELS))
s = Lambda(lambda x: x / 255) (inputs)

c1 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (s)
c1 = Dropout(0.25) (c1)
c1 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c1)
p1 = MaxPooling2D((2, 2)) (c1)

c2 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p1)
c2 = Dropout(0.25) (c2)
c2 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c2)
p2 = MaxPooling2D((2, 2)) (c2)

c3 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p2)
c3 = Dropout(0.5) (c3)
c3 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c3)
p3 = MaxPooling2D((2, 2)) (c3)

c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p3)
c4 = Dropout(0.5) (c4)
c4 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c4)
p4 = MaxPooling2D(pool_size=(2, 2)) (c4)

c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p4)
c5 = Dropout(0.5) (c5)
c5 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', 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='he_normal', padding='same') (u6)
c6 = Dropout(0.5) (c6)
c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c6)

u7 = Conv2DTranspose(512, (2, 2), strides=(2, 2), padding='same') (c6)
u7 = concatenate([u7, c3])
c7 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u7)
c7 = Dropout(0.5) (c7)
c7 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c7)

u8 = Conv2DTranspose(256, (2, 2), strides=(2, 2), padding='same') (c7)
u8 = concatenate([u8, c2])
c8 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u8)
c8 = Dropout(0.25) (c8)
c8 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c8)

u9 = Conv2DTranspose(256, (2, 2), strides=(2, 2), padding='same') (c8)
u9 = concatenate([u9, c1], axis=3)
c9 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u9)
c9 = Dropout(0.25) (c9)
c9 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c9)

cls_depth = Conv2D(1, (1, 1), activation='sigmoid') (c9)


model.compile(optimizer='adam', 
          loss=['mse'],
          metrics=['accuracy'])

准确度始终等于conv2d_19_acc： 0.0015 ..如何改善这一精度？

无法更改批次大小时如何提高训练的准确性

0 个答案: