Dear storleters,
As the deadline approaches I was thinking about the following idea:
End-to-End deep learning with OpenStack Storlets
Imagine that you have a huge dataset from which you could extract
information using machine learning algorithms. The problem is that
datasets usually need to go through a long and tedious curing and
pre-processing before they can be 'presented' to machine learning
algorithm. With large dataset this can get really painful. In this
talk we present how storlets can be used to do an end-to-end
supervised deep learning, thus processing all the data 'in-place'
saving huge amounts of BW. As an example We show face recognition that
starts with off-the-camera jpegs. This involves the following steps:
1. find the face bounding box
2. extract the face part
3. resize to a pre-defined resolution
4. change to greyscale
5. transform into a matrix that can be presented to a learning algorithm
6. train the algorithm over a large training set
We show that all steps can be done using storlets from within a
Jupyter notebook.
Anyone who is interested in taking part please let me know.
Also, this is just an initial suggestion, feel free to suggest other
examples or ideas.
Best,
Eran
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
