Novo Modelo

inputs = Input((IMAGE_HEIGHT, IMAGE_WIDTH, IMAGE_CHANNELS))

s = Lambda(lambda x: x / 255) (inputs)

c1 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (s)
c1 = Dropout(0.1) (c1)
c1 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c1)
p1 = MaxPooling2D((2, 2)) (c1)

c2 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p1)
c2 = Dropout(0.1) (c2)
c2 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c2)
p2 = MaxPooling2D((2, 2)) (c2)

c3 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p2)
c3 = Dropout(0.2) (c3)
c3 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c3)
p3 = MaxPooling2D((2, 2)) (c3)

c4 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p3)
c4 = Dropout(0.2) (c4)
c4 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c4)
p4 = MaxPooling2D(pool_size=(2, 2)) (c4)

c5 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p4)
c5 = Dropout(0.3) (c5)
c5 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c5)
p5 = MaxPooling2D(pool_size=(2, 2)) (c5)

c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p5)
c6 = Dropout(0.3) (c6)
c6 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c6)
p6 = MaxPooling2D(pool_size=(2, 2)) (c6)

c7 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p6)
c7 = Dropout(0.3) (c7)
c7 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c7)
p7 = MaxPooling2D(pool_size=(2, 2)) (c7)

c8 = Conv2D(2048, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p7)
c8 = Dropout(0.3) (c8)
c8 = Conv2D(2048, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c8)
p8 = MaxPooling2D(pool_size=(2, 2)) (c8)

u9 = Conv2DTranspose(1024, (2, 2), strides=(2, 2), padding='same') (c8)
u9 = concatenate([u9, c7])
c9 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u9)
c9 = Dropout(0.2) (c9)
c9 = Conv2D(1024, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c9)

u10 = Conv2DTranspose(512, (2, 2), strides=(2, 2), padding='same') (c9)
u10 = concatenate([u10, c6])
c10 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u10)
c10 = Dropout(0.2) (c10)
c10 = Conv2D(512, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c10)

u11 = Conv2DTranspose(216, (2, 2), strides=(2, 2), padding='same') (c10)
u11 = concatenate([u11, c5])
c11 = Conv2D(216, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u11)
c11 = Dropout(0.2) (c11)
c11 = Conv2D(216, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c11)

u12 = Conv2DTranspose(128, (2, 2), strides=(2, 2), padding='same') (c11)
u12 = concatenate([u12, c4])
c12 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u12)
c12 = Dropout(0.2) (c12)
c12 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c12)

u13 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same') (c12)
u13 = concatenate([u13, c3])
c13 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u13)
c13 = Dropout(0.2) (c13)
c13 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c13)

u14 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same') (c13)
u14 = concatenate([u14, c2])
c14 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u14)
c14 = Dropout(0.2) (c14)
c14 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c14)

u15 = Conv2DTranspose(16, (2, 2), strides=(2, 2), padding='same') (c14)
u15 = concatenate([u15, c1])
c15 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u15)
c15 = Dropout(0.2) (c15)
c15 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c15)

outputs = Conv2D(1, (1, 1), activation='sigmoid') (c15)

model = Model(inputs=[inputs], outputs=[outputs])

#Valor originial do learning_rate = 0.001
#optimizer = tf.keras.optimizers.Adam(lr=0.00001)
#optimizer = tf.keras.optimizers.RMSprop(learning_rate=0.00001)
optimizer = tf.keras.optimizers.SGD(learning_rate=0.00001)

model.compile(optimizer=optimizer, loss='binary_crossentropy',metrics=["accuracy", Precision(name="precision")],)

#model.summary()