Caffe Python 3.6 UnicodeDecodeError: 'ascii' codec can't decode byte 0xe3 in position 5

[tirhum lamauny]

Hello
I use python 3.6 with anaconda 3
I am on CentOS 6, gcc 4.9

I have the following error :

File "make_predictions_1.py", line 51, in <module>
    mean_blob.ParseFromString(f.read())
  File "/work/irlin355_1/gratienj/BigData/local/anaconda3/lib/python3.6/encodings/ascii.py", line 26, in decode
    return codecs.ascii_decode(input, self.errors)[0]
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe3 in position 5: ordinal not in range(128)

when in python script make_predictions_1.py, I try to read the "mean.binaryproto" that have been generate
by the "caffe/bin/compute_image_mean" c++ executable tools that compute the mean image of a collection of pictures.


Does any body have an idea.

I suppose that I have a problem in the python layer that have been generated while compiling caffe framework
In the python module caffe.proto.caffe_pb2, when we read the  mean.binaryproto file , there is a mismatch between
string in ascii or in UTF8 and the transformation in bytes.

I try to regenarate my protoc binary in case it was generated with the wrong python version...

I do not known which error I have done in my caffe framework generation.

I thing that my C++ layer works as the training phase seems OK.
I think I have a problem of compatibility between my C++ layer that generate the mean.binaryproto file
and the python layer that try to read and parse this file.

Thank you for any help or sugestion to check

TiRhum

[tirhum lamauny]

I have found the problem :
With python 3.6, you have to specify 'rb' when you open a file the syntax to read a protobuf file is :

mean_blob = caffe_pb2.BlobProto()
with open('mean.binaryproto','rb') as f :
     mean_blob.ParseFromString(f.read())