Model: "model_1"

Layer (type) Output Shape Param # Connected to

input_1 (InputLayer) (None, 256, 256, 4) 0

conv2d_1 (Conv2D) (None, 126, 126, 16) 1616 input_1[0][0]

leaky_re_lu_1 (LeakyReLU) (None, 126, 126, 16) 0 conv2d_1[0][0]

conv2d_2 (Conv2D) (None, 126, 126, 16) 2320 leaky_re_lu_1[0][0]

conv2d_3 (Conv2D) (None, 126, 126, 16) 6416 leaky_re_lu_1[0][0]

leaky_re_lu_2 (LeakyReLU) (None, 126, 126, 16) 0 conv2d_2[0][0]

leaky_re_lu_3 (LeakyReLU) (None, 126, 126, 16) 0 conv2d_3[0][0]

concatenate_1 (Concatenate) (None, 126, 126, 32) 0 leaky_re_lu_2[0][0]

leaky_re_lu_3[0][0]

conv2d_4 (Conv2D) (None, 61, 61, 32) 25632 concatenate_1[0][0]

leaky_re_lu_4 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_4[0][0]

conv2d_5 (Conv2D) (None, 61, 61, 32) 9248 leaky_re_lu_4[0][0]

conv2d_6 (Conv2D) (None, 61, 61, 32) 25632 leaky_re_lu_4[0][0]

leaky_re_lu_5 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_5[0][0]

leaky_re_lu_6 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_6[0][0]

concatenate_2 (Concatenate) (None, 61, 61, 64) 0 leaky_re_lu_5[0][0]

leaky_re_lu_6[0][0]

conv2d_7 (Conv2D) (None, 61, 61, 32) 2080 concatenate_2[0][0]

leaky_re_lu_7 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_7[0][0]

conv2d_8 (Conv2D) (None, 61, 61, 32) 1056 leaky_re_lu_7[0][0]

conv2d_9 (Conv2D) (None, 61, 61, 32) 9248 leaky_re_lu_7[0][0]

leaky_re_lu_8 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_8[0][0]

leaky_re_lu_9 (LeakyReLU) (None, 61, 61, 32) 0 conv2d_9[0][0]

concatenate_3 (Concatenate) (None, 61, 61, 64) 0 leaky_re_lu_8[0][0]

leaky_re_lu_9[0][0]

conv2d_10 (Conv2D) (None, 29, 29, 64) 102464 concatenate_3[0][0]

leaky_re_lu_10 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_10[0][0]

conv2d_11 (Conv2D) (None, 29, 29, 64) 36928 leaky_re_lu_10[0][0]

conv2d_12 (Conv2D) (None, 29, 29, 64) 102464 leaky_re_lu_10[0][0]

leaky_re_lu_11 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_11[0][0]

leaky_re_lu_12 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_12[0][0]

concatenate_4 (Concatenate) (None, 29, 29, 128) 0 leaky_re_lu_11[0][0]

leaky_re_lu_12[0][0]

conv2d_13 (Conv2D) (None, 29, 29, 64) 8256 concatenate_4[0][0]

leaky_re_lu_13 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_13[0][0]

conv2d_14 (Conv2D) (None, 29, 29, 64) 4160 leaky_re_lu_13[0][0]

conv2d_15 (Conv2D) (None, 29, 29, 64) 36928 leaky_re_lu_13[0][0]

leaky_re_lu_14 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_14[0][0]

leaky_re_lu_15 (LeakyReLU) (None, 29, 29, 64) 0 conv2d_15[0][0]

concatenate_5 (Concatenate) (None, 29, 29, 128) 0 leaky_re_lu_14[0][0]

leaky_re_lu_15[0][0]

conv2d_16 (Conv2D) (None, 13, 13, 64) 204864 concatenate_5[0][0]

leaky_re_lu_16 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_16[0][0]

conv2d_17 (Conv2D) (None, 13, 13, 64) 36928 leaky_re_lu_16[0][0]

conv2d_18 (Conv2D) (None, 13, 13, 64) 102464 leaky_re_lu_16[0][0]

leaky_re_lu_17 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_17[0][0]

leaky_re_lu_18 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_18[0][0]

concatenate_6 (Concatenate) (None, 13, 13, 128) 0 leaky_re_lu_17[0][0]

leaky_re_lu_18[0][0]

conv2d_19 (Conv2D) (None, 13, 13, 64) 8256 concatenate_6[0][0]

leaky_re_lu_19 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_19[0][0]

conv2d_20 (Conv2D) (None, 13, 13, 64) 4160 leaky_re_lu_19[0][0]

conv2d_21 (Conv2D) (None, 13, 13, 64) 36928 leaky_re_lu_19[0][0]

leaky_re_lu_20 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_20[0][0]

leaky_re_lu_21 (LeakyReLU) (None, 13, 13, 64) 0 conv2d_21[0][0]

concatenate_7 (Concatenate) (None, 13, 13, 128) 0 leaky_re_lu_20[0][0]

leaky_re_lu_21[0][0]

conv2d_22 (Conv2D) (None, 11, 11, 64) 73792 concatenate_7[0][0]

leaky_re_lu_22 (LeakyReLU) (None, 11, 11, 64) 0 conv2d_22[0][0]

flatten_1 (Flatten) (None, 7744) 0 leaky_re_lu_22[0][0]

dense_1 (Dense) (None, 512) 3965440 flatten_1[0][0]

leaky_re_lu_23 (LeakyReLU) (None, 512) 0 dense_1[0][0]

dropout_1 (Dropout) (None, 512) 0 leaky_re_lu_23[0][0]

dense_2 (Dense) (None, 12) 6156 dropout_1[0][0]

也不能说复杂，就是层比较多， conv 代表 Convolutional，卷积的意思，在深度学习的体系中全称为Convolutional Neural Network, CNN，卷积神经网络。
总体分为三层 卷积层（图像拆分，特征提取，一般函数命名为 conv相关 ），池化层（特征选择，信息过滤，一般函数命名为pool相关），全连接层（dense 将这些结果进行拟合），dropout是一个参数，其作用是按这个比例丢弃部分结果，防止结果的过拟合（对特定输入拟合结果百分百，换个输入拟合结果很差）,concatenate (将特征进行融合)。
你列举的 这些东西中 (None, 13, 13, 64)，代表的含义对应函数的输入输出格式，这个要对应到函数里看其具体作用，一般conv卷积函数中意味按13，13分割输入，输出64个（实质是对应的数组维度）。 re_lu是激活函数（激活函数有很多种），起作用是将拟合结果变为非线性表达式（更好的进行分类，也更复合实际情况）， flatten 的作用是将多维数组一维化（[[1,2,3],[4,5,6]]-->[1,2,3,4,5,6]）