Binaryproto平均大小与cnn大小不同

Question

我正在尝试使用EmotiW_VGG_S caffemodel CNN。 mean_file binaryproto包含一个3x256x256的图像，我想我必须从要预测的图像中减去。

我获得平均值的方法是：

def binaryproto2npy():
    proto_data = open(mean_file, "rb").read()
    a = caffe.io.caffe_pb2.BlobProto.FromString(proto_data)
    mean_values = caffe.io.blobproto_to_array(a)[0]
    return mean_values

在加载mean_values之后，我尝试将其应用于我的图像（一张脸）并以这种方式对其进行分类：

def process_face(face, mean_values):
    image_dim = 256
    batch_size = 1
    channels = 3
    data = []
    resize = (image_dim, image_dim, 3)

    face = np.resize(face, (face.shape[0], face.shape[1], 1))

    if not ((face.shape[0] == resize[0]) & (face.shape[1] == resize[1])):
        face = caffe.io.resize_image(face, resize)

    # Setting up the right transformer for an input image
    transformer = caffe.io.Transformer({'data': 
    net.blobs['data'].data.shape})
    transformer.set_transpose('data', (2, 0, 1))
    transformer.set_channel_swap('data', (2, 1, 0))
    transformer.set_raw_scale('data', 255.0)
    transformer.set_mean('data', mean_values)

    processed_image = transformer.preprocess('data', face)
    data.append(processed_image)

    net.blobs['data'].reshape(batch_size, channels, 224, 224)
    net.blobs['data'].data[...] = data[0]
    out = net.forward()

问题是我尝试了这段代码，但均值和图像似乎具有不同的形状，因此无法找到解决方案。