how to make an imagenet_mean.binaryproto in python?

[adi]

hi,

??I was wondering how to make an imagenet_mean.binaryproto file on my own dataset while resizing it to 256X256 as it's done in compute_image_mean by using python code

I wrote that code:

imlist =[]

for root,_,files in os.walk(caffe_root+'data/ilsvrc12_vgg/Train'):

  for f in files:

    fullpath=os.path.join(root,f)

    if fullpath[-4:] in [".png",".PNG",".jpg",".JPG",".JPEG"]:#get all the files in an image format into a list 

      imlist.append(fullpath)

#createAvgOfAllImg(imlist)

# Create a numpy array of floats to store the average (assume RGB images)

arr=np.zeros((256,256,3),np.float)

##############################################################################

# Build up average pixel intensities, casting each image as an array of floats

###########################################3##################################

N=len(imlist)

for im in imlist:

    img= Image.open(im).resize(size,Image.ANTIALIAS)

    imarr=np.array(img,dtype=np.float)

    arr=arr+imarr/N

    

newarr=arr

# Round values in array and cast as 8-bit integer

#arr=np.array(np.round(arr),dtype=np.uint8)

######################################################

#making a binaryproto file consists of the mean values

######################################################

imagenet_vgg_mean=caffe_root+'data/ilsvrc12_vgg/imagenet_vgg_mean.binaryproto'

print "MAKING A BINARYPROTO FILE OF MEAN VALUES IN ",imagenet_vgg_mean

blob = caffe.io.array_to_blobproto( arr )

with open( imagenet_vgg_mean, 'wb' ) as f :

    f.write( blob.SerializeToString() )

however, I see that the data in the binaryproto file made by compute_image_mean is different from the one I created..

Am I missing something??

I would like to get a code that created that well..

thanks,