ValueError:应该定义"Dense"输入的最后一个维度.找到了"无"

Question

我的模型定义如下:

def build(data):
    model = Sequential()
    model.add(Cropping2D(cropping=((79, 145), (50, 250)), input_shape= 
                                                                   (160,320,3)))
    model.add(Lambda(lambda x: x/127.5 - 1.0))

    model.add(Conv2D(24, (2, 2), padding='same'))
    model.add(ELU())
    model.add(Conv2D(36, (2, 2), padding='same'))
    model.add(ELU())
    model.add(Conv2D(48, (2, 2), padding='same'))
    model.add(ELU())

    # Add a flatten layer
    model.add(Flatten())
    model.summary()
    model.add(Dense(100))
    model.add(ELU())
    model.add(Dense(50))
    model.add(ELU())
    model.add(Dense(10))
    model.add(ELU())
    model.add(Dense(1))

    return model

得到此错误:

ValueError:Dense应定义输入的最后一个维度.找到了None.

我跑了model.summary(),得到了以下输出

Layer (type)                 Output Shape              Param #   
=================================================================
cropping2d_15 (Cropping2D)   (None, 0, 20, 3)          0         
_________________________________________________________________
lambda_23 (Lambda)           (None, 0, 20, 3)          0         
_________________________________________________________________
conv2d_47 (Conv2D)           (None, 0, 20, 24)         312       
_________________________________________________________________
elu_43 (ELU)                 (None, 0, 20, 24)         0         
_________________________________________________________________
conv2d_48 (Conv2D)           (None, 0, 20, 36)         3492      
_________________________________________________________________
elu_44 (ELU)                 (None, 0, 20, 36)         0         
_________________________________________________________________
conv2d_49 (Conv2D)           (None, 0, 20, 48)         6960      
_________________________________________________________________
elu_45 (ELU)                 (None, 0, 20, 48)         0         
_________________________________________________________________
flatten_12 (Flatten)         (None, None)              0         
=================================================================
Total params: 10,764
Trainable params: 10,764
Non-trainable params: 0

我对python很新,任何输入都会受到赞赏.

Answer 1

您对输入图像的裁剪过多。该cropping论证解释如下：

如果 2 个整数的元组组成的元组：解释为 ((top_crop, Bottom_crop), (left_crop, right_crop))

考虑 Keras 文档中的以下示例：

# Crop the input 2D images or feature maps
model = Sequential()
model.add(Cropping2D(cropping=((2, 2), (4, 4)),
                     input_shape=(28, 28, 3)))
# now model.output_shape == (None, 24, 20, 3)

在您的代码中，您从顶部裁剪 79 像素，从底部裁剪 145 像素，而图像的高度仅为 160 像素。通过较少的裁剪，您的代码运行良好，例如：

model.add(Cropping2D(cropping=((10, 10), (10, 10)), input_shape=(160,320,3)))