blob.channels=0 after reading from mean.binaryproto file

[sgs]

I've created a binaryproto mean file for my dataset. After reading it when I use blob.channels it returns 0. Even blob.height returns 0. But when I use blob.shape it returns-

dim: 1

dim: 3

dim: 400

dim: 300

Which is exactly of same as the numpy array I created the blob from. 

Where am I doing wrong?

I used this code for reading binaryproto file:

import caffe.io

import numpy as np

import sys

if len(sys.argv)!=2:

   print "Too few arguments! Usage: per_channel_mean.py <input_proto_file>"

   sys.exit()

blob = caffe.proto.caffe_pb2.BlobProto()

proto_data = open( sys.argv[1] , 'rb' ).read()

blob.ParseFromString(proto_data)

print blob.channels

This one to create binaryproto: (this creates the file without using lmdb, taking images directly as input)

https://github.com/bamos/python-scripts/blob/aad7a8452d362a249aef95d196595b0b094f9aa5/python2.7/caffe-compute-image-mean.py