These are chat archives for thunder-project/thunder

7th
Feb 2015
Davis Bennett
@d-v-b
Feb 07 2015 01:42
@freeman-lab calling thunder.Colorize.images(np.array([tuneMat[0],tuneMat[1], statsMat]), toType='hsv') returns the following error
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-32-299c138c0048> in <module>()
----> 1 thunder.Colorize.images(np.array([tuneMat[0],tuneMat[1], statsMat]), toType='hsv')

TypeError: unbound method images() must be called with Colorize instance as first argument (got ndarray instance instead)
odds are I'm misunderstanding the usage of thunder.Colorize.images()
ah probably I haven't spawned an instance of thunder.Colorize
Davis Bennett
@d-v-b
Feb 07 2015 01:48
i'm now trying this:
colors = thunder.Colorize(toType='hsv')
img = colors.images(np.array([tuneMat[0], tuneMat[1], statsMat]))
ok this generated an array of the correct size, now I'm not sure how best to visualize it using mpl
calling imshow on slices of the resulting volume is producing black 'images'
Davis Bennett
@d-v-b
Feb 07 2015 01:58
nvm i think the range of values in statsMat is too low, so the images are coming as very dim, probably need to normalize statsMat to [0,1]
Davis Bennett
@d-v-b
Feb 07 2015 04:58
and why might there be NaN's in the output of colors.images?
Jeremy Freeman
@freeman-lab
Feb 07 2015 06:41
@d-v-b yes you need to instantiate the object first. The issue is definitely scale, you can try the scale argument to the Colorize constructor which just amplifies everything. I've been working on a nice auto-scaling behavior for this function but haven't solved it quite yet.
NaNs are probably because there are NaNs in the image, are there?
Davis Bennett
@d-v-b
Feb 07 2015 21:06
not sure how NaNs would have gotten in the image, I'll need to check the output of tuningmodel.fit() and so on