我正在进行一个计算机视觉项目，基于唇部运动的单词分类。有10个类别（单词）需要分类。数据集中每个类别都有一系列图像或帧。我选择了时间分布模型和LSTM模型来完成这个任务。最初，数据集将被转换为numpy数组，并首先输入到CNN层以识别每张图像中的特征。结果随后被输入到时间分布层和LSTM层，以将帧视为时间序列。最后使用了一些密集层进行分类。

我面临的问题是，当我单独训练3到4个类别或单词时，我得到了较高的准确率（约80到90%），并且预测效果非常好。但当我一次性训练10个类别或单词时，准确率非常低。

我不知道这是什么原因。有人能帮我解决这个问题吗？

我的代码

from tensorflow.keras import Sequentialfrom tensorflow.keras.layers import Conv2D, Flatten, Dense, LSTM, Dropout, TimeDistributed,BatchNormalization,MaxPool2D, GlobalMaxPool2Ddef convmodel(shape=(24, 48, 3)):    momentum = .9    model = tf.keras.models.Sequential()    model._name = "CNN1210"        model.add(tf.keras.layers.Conv2D(64, (3,3), input_shape=shape,padding='same', activation='relu', name = "CNN1") )    model.add(tf.keras.layers.Conv2D(64, (3,3), padding='same', activation='relu', name = "CNN2"))    model.add(tf.keras.layers.BatchNormalization(momentum=momentum , name = "Batch1"))        model.add(tf.keras.layers.MaxPool2D(name="Maxpool1") )        model.add(tf.keras.layers.Conv2D(128, (3,3), padding='same', activation='relu', name = "CNN3"))    model.add(tf.keras.layers.Conv2D(128, (3,3), padding='same', activation='relu', name = "CNN4"))    model.add(tf.keras.layers.BatchNormalization(momentum=momentum, name = "batch2") )        model.add(tf.keras.layers.MaxPool2D(name = "Maxpool2"))        model.add(tf.keras.layers.Conv2D(256, (3,3), padding='same', activation='relu', name = "CNN5"))    model.add(tf.keras.layers.Conv2D(256, (3,3), padding='same', activation='relu', name = "CNN6"))    model.add(tf.keras.layers.BatchNormalization(momentum=momentum, name = "Batch3") )        model.add(tf.keras.layers.MaxPool2D(name = "Maxpool3"))        model.add(tf.keras.layers.Conv2D(256, (3,3), padding='same', activation='relu', name = "CNN7"))    model.add(tf.keras.layers.Conv2D(256, (3,3), padding='same', activation='relu', name = "CNN8"))    model.add(tf.keras.layers.BatchNormalization(momentum=momentum, name = "Batch4"))        model.add(tf.keras.layers.MaxPool2D(name = "Maxpool4"))            model.add(tf.keras.layers.Conv2D(512, (3,3), padding='same', activation='relu', name = "CNN9"))    model.add(tf.keras.layers.Conv2D(512, (3,3), padding='same', activation='relu', name = "CNN10"))    model.add(tf.keras.layers.BatchNormalization(momentum=momentum, name = "Batch5"))            # flatten...    model.add(tf.keras.layers.Flatten(name = "Flatten1"))            return modeldef action_model(shapes , nbout=3):    # Create our convnet with (112, 112, 3) input shape    convnet = convmodel(shapes[1:])    print(convnet)    print("convolution over")    # then create our final model    model = tf.keras.models.Sequential()    model._name="1210model"    # add the convnet with (5, 112, 112, 3) shape    model.add(TimeDistributed(convnet, input_shape=shapes, name="Timedist1210"))    print("Time distributed over")    # here, you can also use GRU or LSTM    model.add(tf.keras.layers.LSTM(100, name = "LSTM1210"))    # and finally, we make a decision network    model.add(tf.keras.layers.Dense(1024, activation='relu', name = "Dense12101"))    model.add(tf.keras.layers.Dropout(.8, name = "drop1"))    model.add(tf.keras.layers.Dense(1024, activation='relu', name = "Dense12102"))    model.add(tf.keras.layers.Dropout(.8, name = "drop2"))    model.add(tf.keras.layers.Dense(512, activation='relu', name = "Dense12103"))    model.add(tf.keras.layers.Dropout(.7, name = "drop3"))    model.add(tf.keras.layers.Dense(128, activation='relu', name = "Dense12104"))    model.add(tf.keras.layers.Dropout(.6 ,name = "drop4"))    model.add(tf.keras.layers.Dense(128, activation='relu', name = "Dense12105"))    model.add(tf.keras.layers.Dropout(.5 ,name = "drop5"))    model.add(tf.keras.layers.Dense(64, activation='relu', name = "Dense12106"))    model.add(tf.keras.layers.Dense(32, activation='relu', name = "Dense12107"))    model.add(tf.keras.layers.Dense(16, activation='relu', name = "Dense12108"))    model.add(tf.keras.layers.Dense(8, activation='relu', name = "Dense12109"))    print("Final dense layer")    model.add(tf.keras.layers.Dense(nbout, activation='softmax', name = "Dense12110"))            return modelTimeDistmodel = action_model((10, 24, 48, 3),10)optimizer = tf.keras.optimizers.Adam(0.001)TimeDistmodel.compile(    optimizer,    'categorical_crossentropy',    metrics=['acc'])checkpoint_path = "training_all/cp.ckpt"checkpoint_dir = os.path.dirname(checkpoint_path)# Create a callback that saves the model's weightscp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,                                                 save_weights_only=True,                                                 verbose=1)#TimeDistmodel.summary()finalModel = TimeDistmodel.fit(trainX,trainY,epochs=100, validation_data=(testX,testY),batch_size= 50, callbacks=[cp_callback])

10个类别的输出

Epoch 1/20079/79 [==============================] - ETA: 0s - loss: 2.3026 - acc: 0.1033Epoch 00001: saving model to training_1210\cp.ckpt79/79 [==============================] - 259s 3s/step - loss: 2.3026 - acc: 0.1033 - val_loss: 2.3039 - val_acc: 0.0917Epoch 2/20079/79 [==============================] - ETA: 0s - loss: 2.3029 - acc: 0.1048Epoch 00002: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3029 - acc: 0.1048 - val_loss: 2.3040 - val_acc: 0.0917Epoch 3/20079/79 [==============================] - ETA: 0s - loss: 2.3028 - acc: 0.1038Epoch 00003: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3028 - acc: 0.1038 - val_loss: 2.3040 - val_acc: 0.0917Epoch 4/20079/79 [==============================] - ETA: 0s - loss: 2.3025 - acc: 0.1041Epoch 00004: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3025 - acc: 0.1041 - val_loss: 2.3043 - val_acc: 0.0917Epoch 5/20079/79 [==============================] - ETA: 0s - loss: 2.3025 - acc: 0.0969Epoch 00005: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3025 - acc: 0.0969 - val_loss: 2.3041 - val_acc: 0.0917Epoch 6/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00006: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3043 - val_acc: 0.0917Epoch 7/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1033Epoch 00007: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1033 - val_loss: 2.3044 - val_acc: 0.0917Epoch 8/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00008: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3044 - val_acc: 0.0917Epoch 9/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00009: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3045 - val_acc: 0.0917Epoch 10/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00010: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 11/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00011: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 12/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0985Epoch 00012: saving model to training_1210\cp.ckpt79/79 [==============================] - 243s 3s/step - loss: 2.3024 - acc: 0.0985 - val_loss: 2.3046 - val_acc: 0.0917Epoch 13/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0954Epoch 00013: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0954 - val_loss: 2.3047 - val_acc: 0.0917Epoch 14/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1015Epoch 00014: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1015 - val_loss: 2.3047 - val_acc: 0.0917Epoch 15/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00015: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 16/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0997Epoch 00016: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0997 - val_loss: 2.3047 - val_acc: 0.0917Epoch 17/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00017: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 18/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0974Epoch 00018: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0974 - val_loss: 2.3047 - val_acc: 0.0917Epoch 19/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1000Epoch 00019: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1000 - val_loss: 2.3047 - val_acc: 0.0917Epoch 20/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1023Epoch 00020: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1023 - val_loss: 2.3047 - val_acc: 0.0929Epoch 21/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1028Epoch 00021: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1028 - val_loss: 2.3047 - val_acc: 0.0917Epoch 22/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00022: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 23/20079/79 [==============================] - ETA: 0s - loss: 2.3025 - acc: 0.0990Epoch 00023: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3025 - acc: 0.0990 - val_loss: 2.3048 - val_acc: 0.0917Epoch 24/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00024: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 25/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1005Epoch 00025: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1005 - val_loss: 2.3048 - val_acc: 0.0917Epoch 26/20079/79 [==============================] - ETA: 0s - loss: 2.3025 - acc: 0.0969Epoch 00026: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3025 - acc: 0.0969 - val_loss: 2.3047 - val_acc: 0.0917Epoch 27/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1003Epoch 00027: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1003 - val_loss: 2.3047 - val_acc: 0.0917Epoch 28/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00028: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 29/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00029: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 30/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00030: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 31/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00031: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 32/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0982Epoch 00032: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0982 - val_loss: 2.3046 - val_acc: 0.0917Epoch 33/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00033: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 34/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00034: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 35/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1000Epoch 00035: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1000 - val_loss: 2.3048 - val_acc: 0.0917Epoch 36/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0992Epoch 00036: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0992 - val_loss: 2.3048 - val_acc: 0.0917Epoch 37/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0977Epoch 00037: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0977 - val_loss: 2.3047 - val_acc: 0.0917Epoch 38/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1005Epoch 00038: saving model to training_1210\cp.ckpt79/79 [==============================] - 246s 3s/step - loss: 2.3024 - acc: 0.1005 - val_loss: 2.3047 - val_acc: 0.0917Epoch 39/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00039: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 40/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0941Epoch 00040: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0941 - val_loss: 2.3047 - val_acc: 0.0917Epoch 41/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0980Epoch 00041: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0980 - val_loss: 2.3048 - val_acc: 0.0917Epoch 42/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00042: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 43/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0946Epoch 00043: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0946 - val_loss: 2.3048 - val_acc: 0.0917Epoch 44/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00044: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 45/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0967Epoch 00045: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0967 - val_loss: 2.3048 - val_acc: 0.0929Epoch 46/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0967Epoch 00046: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.0967 - val_loss: 2.3048 - val_acc: 0.0917Epoch 47/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00047: saving model to training_1210\cp.ckpt79/79 [==============================] - 243s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 48/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1010Epoch 00048: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1010 - val_loss: 2.3047 - val_acc: 0.0929Epoch 49/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0995Epoch 00049: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0995 - val_loss: 2.3047 - val_acc: 0.0917Epoch 50/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00050: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 51/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00051: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 52/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1010Epoch 00052: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1010 - val_loss: 2.3046 - val_acc: 0.0917Epoch 53/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00053: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 54/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0954Epoch 00054: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0954 - val_loss: 2.3047 - val_acc: 0.0929Epoch 55/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0992Epoch 00055: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0992 - val_loss: 2.3047 - val_acc: 0.0935Epoch 56/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1026Epoch 00056: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1026 - val_loss: 2.3046 - val_acc: 0.0929Epoch 57/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1031Epoch 00057: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1031 - val_loss: 2.3047 - val_acc: 0.0917Epoch 58/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0987Epoch 00058: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.0987 - val_loss: 2.3047 - val_acc: 0.0917Epoch 59/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0972Epoch 00059: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0972 - val_loss: 2.3047 - val_acc: 0.0929Epoch 60/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1031Epoch 00060: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1031 - val_loss: 2.3047 - val_acc: 0.0917Epoch 61/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00061: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 62/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0957Epoch 00062: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.0957 - val_loss: 2.3047 - val_acc: 0.0917Epoch 63/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00063: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 64/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00064: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 65/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00065: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 66/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00066: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 67/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00067: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 68/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00068: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 69/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1013Epoch 00069: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1013 - val_loss: 2.3047 - val_acc: 0.0917Epoch 70/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1003Epoch 00070: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1003 - val_loss: 2.3047 - val_acc: 0.0917Epoch 71/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0972Epoch 00071: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0972 - val_loss: 2.3047 - val_acc: 0.0917Epoch 72/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0987Epoch 00072: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.0987 - val_loss: 2.3047 - val_acc: 0.0917Epoch 73/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00073: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 74/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00074: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 75/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00075: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 76/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1005Epoch 00076: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1005 - val_loss: 2.3048 - val_acc: 0.0917Epoch 77/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.0964Epoch 00077: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3023 - acc: 0.0964 - val_loss: 2.3048 - val_acc: 0.0917Epoch 78/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00078: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 79/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00079: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 80/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00080: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 81/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00081: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3049 - val_acc: 0.0917Epoch 82/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1013Epoch 00082: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3024 - acc: 0.1013 - val_loss: 2.3046 - val_acc: 0.0917Epoch 83/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00083: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3048 - val_acc: 0.0917Epoch 84/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00084: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 85/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00085: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 86/20079/79 [==============================] - ETA: 0s - loss: 2.3023 - acc: 0.1036Epoch 00086: saving model to training_1210\cp.ckpt79/79 [==============================] - 245s 3s/step - loss: 2.3023 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917Epoch 87/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00087: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3046 - val_acc: 0.0917Epoch 88/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.0962Epoch 00088: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.0962 - val_loss: 2.3047 - val_acc: 0.0917Epoch 89/20079/79 [==============================] - ETA: 0s - loss: 2.3024 - acc: 0.1036Epoch 00089: saving model to training_1210\cp.ckpt79/79 [==============================] - 244s 3s/step - loss: 2.3024 - acc: 0.1036 - val_loss: 2.3047 - val_acc: 0.0917

回答：

我认为这是因为你的训练数据集只有79个样本，太少了。你可以：

1. 通过获取额外的数据来增加训练数据集，例如这个 – http://spandh.dcs.shef.ac.uk/gridcorpus/

或者

2. 使用预训练的权重进行迁移学习，从其他相关的神经网络中获取，例如LipNet – https://github.com/rizkiarm/LipNet。他们的仓库包含了逐步的指导说明。