Question

我试图在谷歌Colaboratory笔记本上用keras / tensorflow进行转学习/瓶颈。我的问题是准确率不超过6％（Kaggle的狗品种挑战，120个类，用datagen.flow_from_directory生成的数据）

以下是我的代码，我有什么遗漏？

tr_model=ResNet50(include_top=False,
                  weights='imagenet',
                 input_shape = (224, 224, 3),)

datagen = ImageDataGenerator(rescale=1. / 255)

#### Training ####
train_generator = datagen.flow_from_directory(train_data_dir,
                                                    target_size=(image_size,image_size),
                                                    class_mode=None,
                                                    batch_size=batch_size,
                                                    shuffle=False)
bottleneck_features_train = tr_model.predict_generator(train_generator)
train_labels = to_categorical(train_generator.classes , num_classes=num_classes)

#### Validation ####
validation_generator = datagen.flow_from_directory(validation_data_dir, 
                                                    target_size=(image_size,image_size),
                                                    class_mode=None,
                                                    batch_size=batch_size,
                                                    shuffle=False)
bottleneck_features_validation = tr_model.predict_generator(validation_generator)
validation_labels = to_categorical(validation_generator.classes, num_classes=num_classes)

#### Model creation ####
model = Sequential()
model.add(Flatten(input_shape=bottleneck_features_train.shape[1:]))
model.add(Dense(num_class, activation='softmax'))

model.compile(optimizer='rmsprop',
              loss='categorical_crossentropy',
              metrics=['accuracy'])

history = model.fit(bottleneck_features_train, train_labels,
                    epochs=30,
                    batch_size=batch_size,
                    validation_data=(bottleneck_features_validation, validation_labels))

我得到val_acc = 0.0592

当我将ResNet50与最后一层一起使用时，得分为82％。

任何人都可以发现我的代码有什么问题。

Answer 1

禁止重新缩放并添加预处理帮助很多。

这些修改极大地帮助了：

from keras.applications.resnet50 import preprocess_input
datagen = ImageDataGenerator(preprocessing_function=preprocess_input)

我现在的准确度为80％

低能力 - 转学习+瓶颈keras-tensorflow（resnet50）

1 个答案: