LSTM理解（附Keras代码实例）

背景：最近在用CRNN做OCR，CRNN的R就是LSTM，从开始深度学习的多年来一直在回避的东西这次躲不过去了，就试着去完整理解一下。

1、我之前的误解。当输入LSTM的Tensor shape为[None, c, n]时，我之前认为会有

c个cell，每一个cell处理一n，这是错的。

      纠正：“时间步”的概念是虚构的，看着图1，仿佛第一个cell，第二个cell，第三个cell都等价并存的，事实上，这是同一个在不同时间的表现。可以这么理解，一个cell是一扇门，过去这扇门的人是数据，门就这么一个，人可以有很多。

图1-经典LSTM示意图

2、 辅证以上说法：LSTM参数计算

输入最后一维是64，输出最后一维是256

计算方法：

4 * (256 * (64 + 256) + 256) = 328704

4：四个门

256、64：输出输入

(64 + 256)：将上一个状态的输出于当前时刻输入合并

最后一个256：bias

3、小问题：t0时刻的上一个时刻的输出是什么？

默认全0 

4、一个CRNN Keras模型代码：

from keras import backend as K
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Input, Dense, Activation
from keras.layers import Reshape, Lambda, BatchNormalization
from keras.layers.merge import add, concatenate
from keras.models import Model
from keras.layers.recurrent import LSTM
# from parameter import *
K.set_learning_phase(0)

# # Loss and train functions, network architecture
def ctc_lambda_func(args):
    y_pred, labels, input_length, label_length = args
    # the 2 is critical here since the first couple outputs of the RNN
    # tend to be garbage:
    y_pred = y_pred[:, 2:, :]
    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)
max_text_len = 7000
num_classes = 7000
# img_w, img_h = 
def get_Model(training):
    input_shape = (None, None, 1)     # (128, 64, 1)

    # Make Networkw
    inputs = Input(name='the_input', shape=input_shape, dtype='float32')  # (None, 128, 64, 1)

    # Convolution layer (VGG)
    inner = Conv2D(64, (3, 3), padding='same', name='conv1', kernel_initializer='he_normal')(inputs)  # (None, 128, 64, 64)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = MaxPooling2D(pool_size=(2, 2), name='max1')(inner)  # (None,64, 32, 64)

    inner = Conv2D(128, (3, 3), padding='same', name='conv2', kernel_initializer='he_normal')(inner)  # (None, 64, 32, 128)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = MaxPooling2D(pool_size=(2, 2), name='max2')(inner)  # (None, 32, 16, 128)

    inner = Conv2D(256, (3, 3), padding='same', name='conv3', kernel_initializer='he_normal')(inner)  # (None, 32, 16, 256)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = Conv2D(256, (3, 3), padding='same', name='conv4', kernel_initializer='he_normal')(inner)  # (None, 32, 16, 256)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = MaxPooling2D(pool_size=(1, 2), name='max3')(inner)  # (None, 32, 8, 256)

    inner = Conv2D(512, (3, 3), padding='same', name='conv5', kernel_initializer='he_normal')(inner)  # (None, 32, 8, 512)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = Conv2D(512, (3, 3), padding='same', name='conv6')(inner)  # (None, 32, 8, 512)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)
    inner = MaxPooling2D(pool_size=(1, 2), name='max4')(inner)  # (None, 32, 4, 512)

    inner = Conv2D(512, (2, 2), padding='same', kernel_initializer='he_normal', name='con7')(inner)  # (None, 32, 4, 512)
    inner = BatchNormalization()(inner)
    inner = Activation('relu')(inner)

    # CNN to RNN
    inner = Reshape(target_shape=((32, 2048)), name='reshape')(inner)  # (None, 32, 2048)
    inner = Dense(64, activation='relu', kernel_initializer='he_normal', name='dense1')(inner)  # (None, 32, 64)

    # RNN layer
    lstm_1 = LSTM(256, return_sequences=True, kernel_initializer='he_normal', name='lstm1')(inner)  # (None, 32, 512)
    lstm_1b = LSTM(256, return_sequences=True, go_backwards=True, kernel_initializer='he_normal', name='lstm1_b')(inner)
    reversed_lstm_1b = Lambda(lambda inputTensor: K.reverse(inputTensor, axes=1)) (lstm_1b)

    lstm1_merged = add([lstm_1, reversed_lstm_1b])  # (None, 32, 512)
    lstm1_merged = BatchNormalization()(lstm1_merged)
    
    lstm_2 = LSTM(256, return_sequences=True, kernel_initializer='he_normal', name='lstm2')(lstm1_merged)
    lstm_2b = LSTM(256, return_sequences=True, go_backwards=True, kernel_initializer='he_normal', name='lstm2_b')(lstm1_merged)
    reversed_lstm_2b= Lambda(lambda inputTensor: K.reverse(inputTensor, axes=1)) (lstm_2b)

    lstm2_merged = concatenate([lstm_2, reversed_lstm_2b])  # (None, 32, 1024)
    lstm2_merged = BatchNormalization()(lstm2_merged)

    # transforms RNN output to character activations:
    inner = Dense(num_classes, kernel_initializer='he_normal',name='dense2')(lstm2_merged) #(None, 32, 63)
    y_pred = Activation('softmax', name='softmax')(inner)

    labels = Input(name='the_labels', shape=[max_text_len], dtype='float32') # (None ,8)
    input_length = Input(name='input_length', shape=[1], dtype='int64')     # (None, 1)
    label_length = Input(name='label_length', shape=[1], dtype='int64')     # (None, 1)

    # Keras doesn't currently support loss funcs with extra parameters
    # so CTC loss is implemented in a lambda layer
    loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc')([y_pred, labels, input_length, label_length]) #(None, 1)

    if training:
        return Model(inputs=[inputs, labels, input_length, label_length], outputs=loss_out)
    else:
        return Model(inputs=[inputs], outputs=y_pred)

model = get_Model(1)
model.summary()

环境：

tf=2.3
keras=2.4.3