python - 喀拉拉邦一类分类

我试图建立一个模型来检测输入图像是否是某个东西（例如，狗或不是）。我在用凯拉斯编码，但准确度太差了。你知道要把这个调准吗？或者我应该使用除keras以外的其他工具来解决一个类分类问题？提前非常感谢。
这是我到目前为止写的代码和输出。

train_dir = './path/to/train_dir'
vali_dir = './path/to/validation_dir'

train_datagen = ImageDataGenerator(
    rescale=1./255,
    rotation_range=40,
    width_shift_range=0.2,
    height_shift_range=0.2,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=False)

test_datagen = ImageDataGenerator(rescale=1./255)

train_generator = train_datagen.flow_from_directory(
        train_dir,
        target_size=(150, 150),
        batch_size=20,
        class_mode='binary')

vali_datagen = ImageDataGenerator(rescale=1./255)

vali_generator = vali_datagen.flow_from_directory(
        vali_dir,
        target_size=(150, 150),
        batch_size=20,
        class_mode='binary')

model = Sequential()
model.add(Conv2D(16, 3, activation='relu', input_shape=(150, 150, 3)))
model.add(MaxPool2D(pool_size=2))
model.add(Conv2D(32, 3, activation='relu'))
model.add(MaxPool2D(pool_size=2))
model.add(Conv2D(64, 3, activation='relu'))
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(Flatten())
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(1, activation='sigmoid'))

model.compile(
        loss='binary_crossentropy',
        optimizer=RMSprop(lr=0.003),
        metrics=['acc']
)

history = model.fit_generator(
        train_generator,
        steps_per_epoch=100,
        epochs=8,
        verbose=2,
        validation_data=vali_generator,
        validation_steps=20
)

输出：

Found 3379 images belonging to 2 classes.
Found 607 images belonging to 2 classes.
Epoch 1/8
 - 136s - loss: 7.6617 - acc: 0.5158 - val_loss: 10.5220 - val_acc: 0.3400
Epoch 2/8
 - 124s - loss: 7.7837 - acc: 0.5118 - val_loss: 10.5220 - val_acc: 0.3400
.......and this is just terrible.

最佳答案

类标签似乎有问题-它们与数据是否正确相关？您可以检查它或发布ImageDataGenerator代码