[CentralOH] focus for numpy/scipy presentations

[Thomas Winningham]

Just wanted to add a note about a thing I was playing with.

Some of the RNN stuff is built around CUDA, but I happened to be on a
machine that didn't have an NVidia card. I wanted to play around with this
"style transfer" stuff. This is sort of a RNN in reverse like Deep Dream
and such, but with regular photos producing a non-LSD like result.

Anyway, I found this library, and after playing with a Docker image that
had an old version of Caffe, I managed to update all of that and get it
working.

https://github.com/fzliu/style-transfer

Mostly all I did was run the demo software that comes with it, but it did
work with just the CPU and no GPU, and could be a fund Saturday afternoon
for anyone on the list to play around with, or send freaky images to
friends or make a new profile picture (heh).

-t

On Mon, May 2, 2016 at 1:08 PM, Neil Ludban <nludban at columbus.rr.com> wrote:

> Eric,
>
> I think this would make a good intro presentation:
> - Load an image from a file into a numpy matrix
> - Display a matrix as an image
> - Basic matrix operations (select a region of interest, convert to
> grayscale)
> - Convolution (smoothing and edge detection filters)
> - Correlation (finding motion between successive images)
>
> As others have pointed out, OpenCV already does this, but for these
> functions it's really just an optimized implementation of simple
> equations (actually, variations on one equation).  You can still
> benefit from using numpy to prepare the inputs and to reduce the
> outputs.  Machine learning on megapixels of input is computationally
> expensive, it is common to preprocess the image in order to reduce
> the amount of time needed.
>
> Are your timelapse images available online, or could you post a small
> number of representative images for me to experiment with?
>
>
> On Sun, 1 May 2016 11:02:20 -0400
> Eric Floehr <eric at intellovations.com> wrote:
> > Neil,
> >
> > I am interested in numpy and scipy for image manipulation and analysis
> and
> > would be interested in getting started with feature detection and image
> > classification.
> >
> > Specifically for my timelapse project, I would like to identify features
> > like birds, the moon, airplane lights, and of course clouds. I would also
> > be interested in grouping sky images into groups automatically. You know
> > that there is blue sky and completely overcast, but there are probably
> > certain other types of sky and cloud cover that naturally group together
> > and it would be neat to be able to identify those clusters.
> >
> > Don't know if any of that is possible, but that's my thoughts :-).
> >
> > Thanks!
> > Eric
> >
> >
> >
> > On Sat, Apr 30, 2016 at 6:27 PM, Neil Ludban <nludban at columbus.rr.com>
> > wrote:
> >
> > > At the last cohpy meeting, several people wrote down numpy and/or scipy
> > > as desired topics for future presentations.  These are very broad and
> > > easily turn into boring overviews of python for matlab people.  Does
> > > anyone have requests for a presentation on a specific scipy module,
> > > digital signal processing topic, or even an idea for a project that you
> > > heard numpy/scipy would be good for but not sure how to get started?
> > > _______________________________________________
> > > CentralOH mailing list
> > > CentralOH at python.org
> > > https://mail.python.org/mailman/listinfo/centraloh
> > >
> _______________________________________________
> CentralOH mailing list
> CentralOH at python.org
> https://mail.python.org/mailman/listinfo/centraloh
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/centraloh/attachments/20160502/52bc6390/attachment.html>