6

我在 python 中使用 caffe 进行分类。我从这里得到代码。在这里，我只使用简单的代码，例如

plt.rcParams['figure.figsize'] = (10, 10)
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
mean_filename='./mean.binaryproto'
proto_data = open(mean_filename, "rb").read()
a = caffe.io.caffe_pb2.BlobProto.FromString(proto_data)
mean = caffe.io.blobproto_to_array(a)[0]
age_net_pretrained='./age_net.caffemodel'
age_net_model_file='./deploy_age.prototxt'
age_net = caffe.Classifier(age_net_model_file, age_net_pretrained,
mean=mean,
channel_swap=(2,1,0),
raw_scale=255,
image_dims=(256, 256))

但是，我遇到了错误，例如

Traceback (most recent call last):
File "cnn_age_gender_demo.py", line 25, in 
image_dims=(256, 256))
File "/home/john/Downloads/caffe/python/caffe/classifier.py", line 34, in init
self.transformer.set_mean(in_, mean)
File "/home/john/Downloads/caffe/python/caffe/io.py", line 255, in set_mean
raise ValueError('Mean shape incompatible with input shape.')
ValueError: Mean shape incompatible with input shape.

你能帮我重新爱上它吗？谢谢

4

4 回答 4

16

将 caffe/python/caffe/io.py 中的第 253-254 行替换掉

if ms != self.inputs[in_][1:]:
    raise ValueError('Mean shape incompatible with input shape.')

经过

if ms != self.inputs[in_][1:]:
    print(self.inputs[in_])
    in_shape = self.inputs[in_][1:]
    m_min, m_max = mean.min(), mean.max()
    normal_mean = (mean - m_min) / (m_max - m_min)
    mean = resize_image(normal_mean.transpose((1,2,0)),in_shape[1:]).transpose((2,0,1)) * (m_max - m_min) + m_min
    #raise ValueError('Mean shape incompatible with input shape.')

重建。希望有帮助

于 2015-06-12T17:31:47.217 回答
1

我遇到了同样的问题，基于 imagenet web demo 我使用这种方式修改了脚本以在第 95 行加载平均文件

mean = np.load(args.mean_file).mean(1).mean(1)

于 2016-09-02T18:11:30.457 回答
0

编辑 deploy_gender.prototxt 并设置： input_dim: 256 input_dim: 256

不知道为什么写错了...

于 2017-11-20T14:43:26.910 回答
0

I am pretty scared to rebuild the code as caffe installation did not come easy for me. But to fix, the solution to resize mean require in_shape (user8264's response), which is set internally in caffe/classifier.py

Anyway, I debugged and found the value for in_shape = (3, 227, 227) for age_net.caffemodel

So the model used for age and gender prediction would the following change:

age_net_pretrained='./age_net.caffemodel'
age_net_model_file='./deploy_age.prototxt'
age_net = caffe.Classifier(age_net_model_file, age_net_pretrained,
                   mean=mean,
                   channel_swap=(2,1,0),
                   raw_scale=255,
                   image_dims=(227, 227))

But mean needs to be modified first: 

m_min, m_max = mean.min(), mean.max()
normal_mean = (mean - m_min) / (m_max - m_min)
in_shape=(227, 227)
mean = caffe.io.resize_image(normal_mean.transpose((1,2,0)),in_shape)
                            .transpose((2,0,1)) * (m_max - m_min) + m_min

This will get rid of "ValueError: Mean shape incompatible with input shape". But I am not sure about the accuracy though. Apparently, for me skipping mean parameter gave better age prediction :)

于 2017-07-22T12:31:23.083 回答