Keras fit_generator和拟合结果不同 [英] Keras fit_generator and fit results are different
问题描述
我正在使用面部图像数据集训练卷积神经网络.数据集包含10,000张尺寸为700 x 700的图像.我的模型有12层.我正在使用生成器函数将图像读取到Keras fit_generator函数中，如下所示.
I am training a Convolutional Neural Network using face images dataset. The dataset has 10,000 images of dimensions 700 x 700. My model has 12 layers. I am using a generator function to read images into Keras fit_generator function as below.
train_file_names ==>包含训练实例文件名的Python列表
train_class_labels ==>一键编码的类标签的整数数组([0，1，0]，[0，0，1]等)
train_data ==>大量的训练实例
train_steps_epoch ==> 16(批处理大小为400，我有6400个要训练的实例.因此，一次遍历整个数据集需要16次迭代)
batch_size ==> 400
calls_made ==>当生成器到达训练实例的末尾时，它将重置索引以在下一个时期从第一个索引加载数据.
train_file_names ==> Python list containing filenames of training instances
train_class_labels ==> Numpy array of one-hot encoded class lables ([0, 1, 0], [0, 0, 1] etc.)
train_data ==> Numpy array of training instances
train_steps_epoch ==> 16 (Batch size is 400 and I have 6400 instances for training. Hence it takes 16 iterations for a single pass through the whole dataset)
batch_size ==> 400
calls_made ==> When generator reaches end of training instances, it resets indexes to load data from first index in next epoch.
我正在将此生成器作为参数传递给keras'fit_generator'函数，以为每个时期生成新的一批数据.
I am passing this generator as an argument to keras 'fit_generator' function to generate new batch of data for each epoch.
val_data，val_class_labels ==>验证数据Numpy数组
纪元==>纪元数
val_data, val_class_labels ==> Validation data numpy arrays
epochs ==> No. of epochs
使用Keras fit_generator :
model.fit_generator(generator=train_generator, steps_per_epoch=train_steps_per_epoch, epochs=epochs, use_multiprocessing=False, validation_data=[val_data, val_class_labels], verbose=True, callbacks=[history, model_checkpoint], shuffle=True, initial_epoch=0)
代码
def train_data_generator(self):
index_start = index_end = 0
temp = 0
calls_made = 0
while temp < train_steps_per_epoch:
index_end = index_start + batch_size
for temp1 in range(index_start, index_end):
index = 0
# Read image
img = cv2.imread(str(TRAIN_DIR / train_file_names[temp1]), cv2.IMREAD_GRAYSCALE).T
train_data[index] = cv2.resize(img, (self.ROWS, self.COLS), interpolation=cv2.INTER_CUBIC)
index += 1
yield train_data, self.train_class_labels[index_start:index_end]
calls_made += 1
if calls_made == train_steps_per_epoch:
index_start = 0
temp = 0
calls_made = 0
else:
index_start = index_end
temp += 1
gc.collect()
fit_generator的输出
Epoch 86/300
16/16 [==============================]-16s 1s/step-损失:1.5739-acc:0.2991-val_loss :12.0076-val_acc:0.2110
时代87/300
16/16 [==============================]-16s 1s/step-损失:1.6010-acc:0.2549-val_loss :11.6689-val_acc:0.2016
时代88/300
16/16 [==============================]-16s 1s/step-损失:1.5750-acc:0.2391-val_loss :10.2663-val_acc:0.2004
时代89/300
16/16 [==============================]-16s 1s/step-损失:1.5526-acc:0.2641-val_loss :11.8809-val_acc:0.2249
时代90/300
16/16 [==============================]-16s 1s/step-损失:1.5867-acc:0.2602-val_loss :12.0392-val_acc:0.2010
时代91/300
16/16 [==============================]-16s 1s/step-损失:1.5524-acc:0.2609-val_loss :12.0254-val_acc:0.2027
Epoch 86/300
16/16 [==============================] - 16s 1s/step - loss: 1.5739 - acc: 0.2991 - val_loss: 12.0076 - val_acc: 0.2110
Epoch 87/300
16/16 [==============================] - 16s 1s/step - loss: 1.6010 - acc: 0.2549 - val_loss: 11.6689 - val_acc: 0.2016
Epoch 88/300
16/16 [==============================] - 16s 1s/step - loss: 1.5750 - acc: 0.2391 - val_loss: 10.2663 - val_acc: 0.2004
Epoch 89/300
16/16 [==============================] - 16s 1s/step - loss: 1.5526 - acc: 0.2641 - val_loss: 11.8809 - val_acc: 0.2249
Epoch 90/300
16/16 [==============================] - 16s 1s/step - loss: 1.5867 - acc: 0.2602 - val_loss: 12.0392 - val_acc: 0.2010
Epoch 91/300
16/16 [==============================] - 16s 1s/step - loss: 1.5524 - acc: 0.2609 - val_loss: 12.0254 - val_acc: 0.2027
我的问题是，在将"fit_generator"与上述生成器函数一起使用时，我的模型损失根本没有改善，验证准确性也很差.但是，当我按如下方式使用keras的拟合"函数时，模型损失会减少，验证准确性会更好.
My problem is, while using 'fit_generator' with above generator function as above, my model loss is not at all improving and validation accuracy is very poor. But when I use keras 'fit' function as below, the model loss decreases and validation accuracy is far better.
使用Keras拟合函数而不使用生成器
model.fit(self.train_data, self.train_class_labels, batch_size=self.batch_size, epochs=self.epochs, validation_data=[self.val_data, self.val_class_labels], verbose=True, callbacks=[history, model_checkpoint])
使用拟合函数训练后的输出
Epoch 25/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0207-acc:0.9939-val_loss :4.1009-val_acc:0.4916
时代26/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0197-acc:0.9948-val_loss :2.4758-val_acc:0.5568
时代27/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0689-acc:0.9800-val_loss :1.2843-val_acc:0.7361
时代28/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0207-acc:0.9947-val_loss :5.6979-val_acc:0.4560
时代29/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0353-acc:0.9908-val_loss :1.0801-val_acc:0.7817
时代30/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0362-acc:0.9896-val_loss :3.7851-val_acc:0.5173
时代31/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0481-acc:0.9896-val_loss :1.1152-val_acc:0.7795
时代32/300
6400/6400 [==============================]-20s 3ms/step-损耗:0.0106-acc:0.9969-val_loss :1.4803-val_acc:0.7372
Epoch 25/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0207 - acc: 0.9939 - val_loss: 4.1009 - val_acc: 0.4916
Epoch 26/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0197 - acc: 0.9948 - val_loss: 2.4758 - val_acc: 0.5568
Epoch 27/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0689 - acc: 0.9800 - val_loss: 1.2843 - val_acc: 0.7361
Epoch 28/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0207 - acc: 0.9947 - val_loss: 5.6979 - val_acc: 0.4560
Epoch 29/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0353 - acc: 0.9908 - val_loss: 1.0801 - val_acc: 0.7817
Epoch 30/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0362 - acc: 0.9896 - val_loss: 3.7851 - val_acc: 0.5173
Epoch 31/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0481 - acc: 0.9896 - val_loss: 1.1152 - val_acc: 0.7795
Epoch 32/300
6400/6400 [==============================] - 20s 3ms/step - loss: 0.0106 - acc: 0.9969 - val_loss: 1.4803 - val_acc: 0.7372
推荐答案
您必须确保数据生成器在各个时期之间对数据进行混洗.我建议您在循环外部创建一个可能的索引列表，使用random.shuffle将其随机化，然后在循环内部对其进行迭代.
You have to make sure that your data generator shuffles the data between epochs. I would suggest you create a list of possible indices outside of your loop, randomize it with random.shuffle and then iterate over that inside your loop.
来源: https://github.com/keras-team/keras/issues /2389 和自己的经验.
这篇关于Keras fit_generator和拟合结果不同的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
