I'm following along with the Caffe/Python example from http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/00-classification.ipynb
Everything runs fine until I get to this line in step 2:transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
... which throws an Exception:
Traceback (most recent call last):
File "caffe_python.py", line 43, in <module>
transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
File "/ad/eng/opt/64/caffe-rc3/python/caffe/pycaffe.py", line 28, in _Net_blobs
return OrderedDict(zip(self._blob_names, self._blobs))
TypeError: No to_python (by-value) converter found for C++ type: boost::shared_ptr<caffe::Blob<float> >
Any ideas? Looking at earlier threads I see a few different possible problems but nothing definitive. I thought it was possible something wrong with my Boost.Python install, but I built and tested some examples with no trouble. The caffe python module itself imports just fine, and other calls in that tutorial work OK; it's not until that specific line that there's a problem.
I'm using:
CentOS 6.8
Caffe rc3
Boost 1.60.0
Anaconda Python 2.7
Thanks for any thoughts!