Clean and simple Keras implementation of the residual block (non-bottleneck) accompanying Deep Residual Learning: https://blog.waya.ai/deep-residual-learning-9610bb62c355.

residual_block.py

from keras import layers


def residual_block(y, nb_channels, _strides=(1, 1), _project_shortcut=False):
    shortcut = y

    # down-sampling is performed with a stride of 2
    y = layers.Conv2D(nb_channels, kernel_size=(3, 3), strides=_strides, padding='same')(y)
    y = layers.BatchNormalization()(y)
    y = layers.LeakyReLU()(y)

    y = layers.Conv2D(nb_channels, kernel_size=(3, 3), strides=(1, 1), padding='same')(y)
    y = layers.BatchNormalization()(y)

    # identity shortcuts used directly when the input and output are of the same dimensions
    if _project_shortcut or _strides != (1, 1):
        # when the dimensions increase projection shortcut is used to match dimensions (done by 1×1 convolutions)
        # when the shortcuts go across feature maps of two sizes, they are performed with a stride of 2
        shortcut = layers.Conv2D(nb_channels, kernel_size=(1, 1), strides=_strides, padding='same')(shortcut)
        shortcut = layers.BatchNormalization()(shortcut)

    y = layers.add([shortcut, y])
    y = layers.LeakyReLU()(y)

    return y

How would i add it to my model?

Thanks for providing that clean implementation. 😊 I always add the "use_bias=False" argument right before a batchnorm layer. Do you have any reasons to use bias in Conv2D before batchnorm?

How would i add it to my model?

@MarviB16 I have a similar code and you can add it to your model this way

i = Input(shape=(x_train.shape[1],x_train.shape[2],x_train.shape[3]))
d = Conv2D(64,(7,7),activation='elu',kernel_regularizer=regularizers.l2(0.0001),padding='same')(i)#Con esto llegue a 0.716
d = residual_block(i,64,128,(1,1),False)# 64 ,64 para 0.716 #64 ,128 para 0.72 #ERA FALSE PARA 0.72
d = MaxPooling2D(pool_size=(2, 2))(d)
d = GaussianNoise(0.33)(d)#0.33 con 0.716
d = residual_block(d,128,256,sizematches=True)# 128,256 para 0.72 #ERA FALSE PARA 0.72
d = MaxPooling2D(pool_size=(2, 2))(d)
d = GaussianNoise(0.43)(d)#0.43 con 0.716
d = residual_block(d,256,512,sizematches=True) #ERA FALSE PARA 0.72
d = GaussianNoise(0.53)(d)#0.53 para 0.716
d = AveragePooling2D(pool_size=(4, 4))(d)#0.716 con pool size 4,4
d = Flatten()(d)
d= Dense(y_train.shape[1],activation='softmax')(d)