Mailman 3 July 2013 - scikit-image

Re: skimage and opencv
by Emmanuelle Gouillart 22 Sep '16

22 Sep '16

Hi FranÃ¯Â¿Â½ois, that's an excellent question, and not a troll :-). Opencv is a very powerful library, but it focuses primarily on computer vision (feature detection and extraction, classification, ...), as opposed to image processing in general (with other tasks such as denoising, segmentation, ...). The other big difference is that skimage builds on numpy ndarrays, and uses the full power of the numpy API (including of course the basic facilities for processing arrays as images that come with numpy), as well as some of scipy functions (you could have added scipy.ndimage to your list -- a few functions in skimage are wrappers around scipy.ndimage, that exist for the sake of completeneness). One important consequence is that algorithms working for 3-d or even n-d images can be easily implemented in 3-d/n-d in skimage, whereas opencv is restricted to 2-D images (as far as I know). Thanks to the use of numpy arrays, the API of skimage is also quite pleasant for a numpy user, more than the API of opencv. A related difference is that skimage is written in python and cython, whereas opencv is a C++ library. The two libraries attract a different crowd of developers, and a Python/Cython toolkit based on numpy arrays is easier to develop and maintain inside the Scientific Python ecosystem. I'm sure that other devs/users will have things to add to this discussion! Cheers, Emmanuelle On Thu, Dec 27, 2012 at 02:06:08PM -0800, FranÃ¯Â¿Â½ois wrote: > Hi users and devs, > It came to my knowledge that another python library (based on C++ and C > codes) for image processing exists too : opencv > I understand that numpy intregrates some basic features and we need some > advanced features but I have the feeling that skimages is redoundant with > opencv in some ways. > What's the position of skimage about that? (Don't read this question as a > troll but like a real question). > I mean that similar features exist in both. Would not be possible to > reuse/integrate opencv or merge? what's the reason for keeping them apart? > My observation is there is 4 libraries to manipulate images: > * PIL > * numpy > * skimages > * opencv > That's a lot. > Cheers,

3 2

Re: Contribution: circles and hough transform
by François 18 Jun '14

18 Jun '14

Hi, FYI, I plan to work on the detection of ellipses quite soon. Stefan indicated to me a reference which seems fairly simple to implement. Cheers, François.

3 3

Get border pixels of labelled region
by Robin Wilson 12 Sep '13

12 Sep '13

Hi all, Is there a way to get the pixels on the border of a labelled region in an image using skimage? I've done edge detection, and then filled the resulting shapes (after some edge linking), and would now like to be able to get the pixels on the border of each of these shapes. I would have thought that getting borders would be an option in regionprops, but it doesn't seem to be. Any ideas? Cheers, Robin

3 4

Sift on GPU
by Jérôme Kieffer 10 Sep '13

10 Sep '13

Dear Pythonistas, We are porting the SIFT keypoints extraction algorithm (available from IPOL) to GPU using PyOpenCL. For the moment, the keypoint location works and shows a speed-up of 5 to 10x (without tuning so far, vs C++). A lot of work is remaining, especially: * limit the memory footprint (700MB/10Mpix image currently) * calculate the descriptor for each descriptor * keypoint matching and image alignment. * best interleave of IO/CPU/GPU but we managed to port the most trickiest part to OpenCL (without using textures, which makes it running also on multi-core). I would like to thank the people who published their algorithm on IPOL; making unit testing possible. Last but not least, the code is open source and should have a BSD licence (even if there is a patent on the algorithm in the USA). https://github.com/pierrepaleo/sift_pyocl Cheers, -- Jérôme Kieffer <google(a)terre-adelie.org>

4 5

random walker segmentation
by Emmanuelle Gouillart 05 Aug '13

05 Aug '13

Hello, a while ago, I contributed to skimage an implementation of the random walker segmentation algorithm (which has been improved and extended by many others since then). This algorithm computes a multilabel segmentation using seeds (already labeled pixels), by determining for an unlabeled pixel the probability that a seed diffuses to the pixel (with an anisotropic diffusion coefficient depending on the gradient between neighboring pixels). In the current implementation in skimage, the computation of the probability map is done by inverting a large sparse linear system (involving the Laplacian of the graph of pixels). Different methods can be chosen to solve the linear system: a brute force inversion only works for tiny images; a conjugate gradient method works well but is quite slow. If the package pyamg is available, a multigrid method is used to compute a preconditioner, which speeds up the algorithm -- but it requires pyamg. Also, the memory cost of the algorithm is important (linear, I think, though. I haven't yet taken the time to use a memory profiler but I should). Recently, a colleague brought to my attention that the linear system was just a set of fixed-point equations, that could be solved iteratively. Indeed, the solution verifies that the probability of a pixel is the weighted sum (with weights on edges that are a decreasing function of gradients) of the probabilities of its neighbors. I have written a quick and dirty implementation (only for 2-D grayscale images and for 2 labels) of this "local" version, available on https://github.com/emmanuelle/scikits.image/blob/local_random_walker/skimag… It turns out that this implementation is slower than the conjugate gradient with multigrid acceleration (typically 2 to three times slower), but it has several advantages. First, it can be as fast as the "simple" conjugate gradient (without pyamg's muligrid acceleration), which is the mode that most users will use (we don't expect users to install pymag when they are just trying out algorithms). Second, its memory cost is lower (for example, the weight of an edge is stored only once, while it appears twice in the Laplacian matrix). Finally, because the operations only involve neighboring pixels, it is possible that further speed optimization can be done (using cython... or maybe a GPU implementation, even if we're not that far yet with skimage). So, should we replace the linear algebra implementation with this simpler local and iterative implementation ? I'd be interested in knowing about your opinion. Cheers, Emmanuelle

3 3

label
by Evan Daugharthy 31 Jul '13

31 Jul '13

Hello all, Based on the discussion here: https://groups.google.com/forum/#!topic/pythonvision/AVrnueiKKYI It seems like this label function should be faster than the scipy version. However, I find it to be much slower: from skimage import morphology from scipy import ndarray as nd k = np.zeros((4000,4000),dtype=int) k[100,100] = 1 strel = np.ones((3,3)) %timeit morphology.label(k) 1 loops, best of 3: 1.46 s per loop %timeit nd.label(k,strel) 1 loops, best of 3: 355 ms per loop If you have any insight, I would be very grateful. I am using version 0.8.2 on the versions of python/numpy/cython, etc. included in the most recent edition of Enthought/Canopy. Thanks, Evan Daugharthy

1 0

Re: Blog updated
by Chintak Sheth 31 Jul '13

31 Jul '13

Thanks a lot Ankit! =) Yeah, I beginning to really like NumPy! :D Chintak

1 0

Re: Blog updated
by Ankit Agrawal 31 Jul '13

31 Jul '13

Great posts Chintak !! They cleared up a lot of the ambiguities I had about ndarray's working at a lower level of abstraction. Same goes for Stefan's clear 'The Numpy Array' paper. Thanks a lot. Ankit Agrawal, Communication and Signal Processing, IIT Bombay. On Wed, Jul 31, 2013 at 4:25 PM, Chintak Sheth <chintaksheth(a)gmail.com>wrote: > Added the explanation and another example in einsum section. Hope that > makes it more understandable. Thanks for pointing it out, Juan! :D > > @Stefan, sure I'll add it in a bit. I'm trying to get Python3 up and > rolling as you had instructed. > > Chintak > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Re: Blog updated
by Chintak Sheth 31 Jul '13

31 Jul '13

Added the explanation and another example in einsum section. Hope that makes it more understandable. Thanks for pointing it out, Juan! :D @Stefan, sure I'll add it in a bit. I'm trying to get Python3 up and rolling as you had instructed. Chintak

1 0

Re: Contiguous Array
by Juan Nunez-Iglesias 31 Jul '13

31 Jul '13

@Marc: They are very different functions. ravel() gives you a view or a copy of the array as a 1D array that happens to be contiguous. ascontiguousarray will give you a view or a copy of the array *of the same shape* as the original, but guaranteed to be contiguous. Let me try to illustrate: In [1]: x = np.arange(20, dtype=np.uint8).reshape((4,5)) In [2]: x Out[2]: array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]], dtype=uint8) In [3]: x.strides Out[3]: (5, 1) So I've made an array x that has 4 rows and 5 columns. Internally, the data in x is stored as a contiguous block of memory of length 20 bytes. The strides tell numpy that to get to the next row in the array, you need to skip 5 bytes in the memory, but to get to the next column, you only need to go to the next (+1) byte in memory. This is called C-contiguous. Now, let's get a weird view of x: In [4]: y = x.T[1::2] I've taken the transpose of x (5 rows and 4 columns) and then taken the 1st and 3rd rows of that. By default, numpy will use nice tricks with strides to avoid a copy: In [5]: y.strides Out[5]: (2, 5) In [6]: y Out[6]: array([[ 1, 6, 11, 16], [ 3, 8, 13, 18]], dtype=uint8) In [7]: y[1, 1] = 45 In [8]: x Out[8]: array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 45, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]], dtype=uint8) So, y is actually pointing to the same memory as x, but it uses different strides to access its own rows and columns. Notice that changing an entry in y changed the corresponding entry in x. In [10]: np.ravel(y) Out[10]: array([ 1, 6, 11, 16, 3, 45, 13, 18], dtype=uint8) np.ravel gives you a "linearised" version of y, concatenating all the rows together. In [11]: z = np.ascontiguousarray(y) In [12]: z Out[12]: array([[ 1, 6, 11, 16], [ 3, 45, 13, 18]], dtype=uint8) In [13]: z.strides Out[13]: (4, 1) np.ascontiguousarray gives you the same thing as y, except now it's a separate contiguous block of memory: look at the strides. You can tell that it's a copy because modifying z has no effect on x: In [14]: z[0, 0] = 72 In [15]: x Out[15]: array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 45, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]], dtype=uint8) In [16]: z Out[16]: array([[72, 6, 11, 16], [ 3, 45, 13, 18]], dtype=uint8) But, if an array is already contiguous, np.ravel() avoids a copy and returns a view into the same data. So: In [19]: a = np.ravel(x) In [20]: a[0] = 72 In [21]: x Out[21]: array([[72, 1, 2, 3, 4], [ 5, 6, 7, 45, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]], dtype=uint8) If you try to get a contiguous version of x, which is already contiguous, you are actually getting x back: In [22]: b = np.ascontiguousarray(x) In [23]: b is x Out[23]: True In [25]: b[3, 4] = 0 In [26]: x Out[26]: array([[72, 1, 2, 3, 4], [ 5, 6, 7, 45, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 0]], dtype=uint8) Hope that clarifies a few things! It takes quite a bit of playing around before you can get an intuition for what's a copy, what's a view, what's contiguous, etc. On Wed, Jul 31, 2013 at 6:15 PM, Ankit Agrawal <aaaagrawal(a)gmail.com> wrote: > Hi Marc, > > > On Wed, Jul 31, 2013 at 1:09 PM, Marc de Klerk <deklerkmc(a)gmail.com>wrote: > >> Hi guys, >> >> I've been using np.ravel(). This morning I tried to lookup the difference >> between np.ravel() and np.ascontiguousarray(). Does anybody know? >> > I am not sure if this helps as I don't know your purpose for using > np.ravel / np.ascontiguousarray. I got to know about the ndarray.flags > method yesterday from Stefan while discussion on this PR<https://github.com/scikit-image/scikit-image/pull/668> > . > > In [15]: a = np.arange(20).reshape((4,5)) > > In [16]: a > Out[16]: > array([[ 0, 1, 2, 3, 4], > [ 5, 6, 7, 8, 9], > [10, 11, 12, 13, 14], > [15, 16, 17, 18, 19]]) > > In [17]: a.flags > Out[17]: > C_CONTIGUOUS : True > F_CONTIGUOUS : False > OWNDATA : False > WRITEABLE : True > ALIGNED : True > UPDATEIFCOPY : False > > In [18]: b = np.ravel(a) > > In [20]: b > Out[20]: > array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, > 17, 18, 19]) > > In [21]: b.flags > Out[21]: > C_CONTIGUOUS : True > F_CONTIGUOUS : True > OWNDATA : False > WRITEABLE : True > ALIGNED : True > UPDATEIFCOPY : False > > Hope this helps!! > > Marc >> >> On Sunday, July 21, 2013 6:37:47 AM UTC+2, Chintak Sheth wrote: >> >>> Hi Ronnie, >>> >>> On Jul 21, 2013 10:00 AM, "Ronnie Ghose" <ronnie...(a)gmail.com> wrote: >>> > >>> > So in skimage/colors why does it matter if the array is contiguous? Is >>> this for Cython operations later? >>> > >>> >>> Yeah it is mainly for using memory views in Cython which is initialized >>> as C contiguous. >>> `cdef some_type[:. ::1] var_name` >>> >>> In thus case ::1 is for C contiguous. >>> >>> Chintak >>> >> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Re: Blog updated
by Juan Nunez-Iglesias 31 Jul '13

31 Jul '13

Yep, very cool posts Chintak! Especially the second one where you turn it up to 11! =) And it really shows the power of numpy, beyond what I realised! On Wed, Jul 31, 2013 at 6:23 AM, Stéfan van der Walt <stefan(a)sun.ac.za>wrote: > The part about einsum is really nice! I propose we add a nice comment > in the code (of your PR) to explain what's happening on that einsum > line, because most people won't have a clue :) > This is definitely true of me! Actually I don't really understand the "real" einsum expression in your blog post, the one doing the SSD computation. Can you explain it in a bit more detail? What's the arrow? Anyway, good stuff. =)

1 0

Re: Contiguous Array
by Marc de Klerk 31 Jul '13

31 Jul '13

Hi guys, I've been using np.ravel(). This morning I tried to lookup the difference between np.ravel() and np.ascontiguousarray(). Does anybody know? Marc On Sunday, July 21, 2013 6:37:47 AM UTC+2, Chintak Sheth wrote: > > Hi Ronnie, > > On Jul 21, 2013 10:00 AM, "Ronnie Ghose" <ronnie...(a)gmail.com<javascript:>> > wrote: > > > > So in skimage/colors why does it matter if the array is contiguous? Is > this for Cython operations later? > > > > Yeah it is mainly for using memory views in Cython which is initialized as > C contiguous. > `cdef some_type[:. ::1] var_name` > > In thus case ::1 is for C contiguous. > > Chintak >

3 2

Re: Blog updated
by Stéfan van der Walt 30 Jul '13

30 Jul '13

Hi Chintak On Tue, Jul 30, 2013 at 10:12 PM, Chintak Sheth <chintaksheth(a)gmail.com> wrote: > https://chintaksheth.wordpress.com/2013/07/31/numpy-the-tricks-of-the-trade… The part about einsum is really nice! I propose we add a nice comment in the code (of your PR) to explain what's happening on that einsum line, because most people won't have a clue :) You can often write einsum as some operation + a reduction over one or more axes. Can this operation be expressed in both ways? If so, which do you find to be more readable and understandable? Thanks! Stéfan

1 0

Blog updated
by Chintak Sheth 30 Jul '13

30 Jul '13

Hi, Blog updated. Any feedback, welcome! https://chintaksheth.wordpress.com/2013/07/30/numpy-the-tricks-of-the-trade… https://chintaksheth.wordpress.com/2013/07/31/numpy-the-tricks-of-the-trade… Chintak

1 0

Request for review: #654
by jostein.floystad＠gmail.com 28 Jul '13

28 Jul '13

Dear core devs, I'd like to ask for review on this PR, https://github.com/scikit-image/scikit-image/pull/654 which will hopefully finish the clean-up of the Radon transform module. Kind regards, Jostein

1 0

euroscipy sprint
by Emmanuelle Gouillart 28 Jul '13

28 Jul '13

Hi all, do we plan to have a scikit-image sprint at Euroscipy in Brussels? There is one day for sprints, Sunday August 25. I'll be there! Anyone else? Cheers, Emmanuelle

3 2

Re: Error executing plot_circular_hough_transform.py
by Ravi Prakash 28 Jul '13

28 Jul '13

Hi Stéfan, I tried with the latest version. It worked fine. Thank you. Hi Tony, I did not understand what r's value should be. Could you please elaborate? I am new to skikit-image as well as numpy. Ravi On Saturday, July 27, 2013 6:10:50 AM UTC+5:30, Tony S Yu wrote: > > > On Jul 26, 2013 8:52 AM, "Stéfan van der Walt" <ste...(a)sun.ac.za<javascript:>> > wrote: > > > > Hi Ravi > > > > On Fri, Jul 26, 2013 at 3:13 PM, Ravi Prakash <praks.p...(a)gmail.com<javascript:>> > wrote: > > > Traceback (most recent call last): > > > File "plot_circular_hough_transform.py", line 69, in <module> > > > image[cy, cx] = (220, 20, 20) > > > IndexError: index 102 is out of bounds for axis 0 with size 95 > > > > > > I am at version skimage 0.8.2, Python 2.7.2+, os: Linux Mint 12 Lisa. > > > > Could you try with the latest version from GitHub? It might have been > > a bug in 0.8.2. > > > > Stéfan > > > > -- > > Just to expand on Stefan's reply: the old implementation (0.8) searched > for any part of the circle in the image, so the center could as much as "r" > outside the image. You can convert the 0.8 result to the 0.9 result with > > hough_result = hough_result[r:-r, r:-r] > > Best, > Tony > > > You received this message because you are subscribed to the Google > Groups "scikit-image" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to scikit-image...(a)googlegroups.com <javascript:>. > > For more options, visit https://groups.google.com/groups/opt_out. > > > > >

2 2

Re: Error executing plot_circular_hough_transform.py
by Tony Yu 27 Jul '13

27 Jul '13

On Jul 26, 2013 8:52 AM, "Stéfan van der Walt" <stefan(a)sun.ac.za> wrote: > > Hi Ravi > > On Fri, Jul 26, 2013 at 3:13 PM, Ravi Prakash <praks.putchala(a)gmail.com> wrote: > > Traceback (most recent call last): > > File "plot_circular_hough_transform.py", line 69, in <module> > > image[cy, cx] = (220, 20, 20) > > IndexError: index 102 is out of bounds for axis 0 with size 95 > > > > I am at version skimage 0.8.2, Python 2.7.2+, os: Linux Mint 12 Lisa. > > Could you try with the latest version from GitHub? It might have been > a bug in 0.8.2. > > Stéfan > > -- Just to expand on Stefan's reply: the old implementation (0.8) searched for any part of the circle in the image, so the center could as much as "r" outside the image. You can convert the 0.8 result to the 0.9 result with hough_result = hough_result[r:-r, r:-r] Best, Tony > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > >

1 0

How to get a region chain code?
by Valeriy Sokolov 26 Jul '13

26 Jul '13

Hi, I am studying a shape descriptors and would like to compute a chain code description of a labeled shape. I was trying to look into docs but have not find a direct way to get this region property with skimage. Is there any way to do it in a few calls which I missed? Thank you in advance.

5 4

Error executing plot_circular_hough_transform.py
by Ravi Prakash 26 Jul '13

26 Jul '13

Hi, When I execute plot_circular_hough_transform.py it gives the following error: Traceback (most recent call last): File "plot_circular_hough_transform.py", line 69, in <module> image[cy, cx] = (220, 20, 20) IndexError: index 102 is out of bounds for axis 0 with size 95 I am at version skimage 0.8.2, Python 2.7.2+, os: Linux Mint 12 Lisa. Could you please help me execute this? Please let me know if you need more information about my execution environment. Thank you. Ravi

2 1

Re: Welcome to the core team, @jni!
by Johannes Schönberger 26 Jul '13

26 Jul '13

Hi, Sorry for the delayed welcome from my side... Great to have you in the team! Johannes Schönberger Am 23.07.2013 um 13:29 schrieb "Stéfan van der Walt" <stefan(a)sun.ac.za>: > Dear all, > > It gives me great pleasure to welcome Juan Nunez-Iglesias to the core > skimage team! Juan certainly earned his way by suffering through many > of our most nit picky reviews, and through his excellent > contributions, including the recent N-d extension to SLIC > segmentation. > > Juan, it's great to have you on board. > > Stéfan > > -- > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > >

1 0

Re: Welcome to the core team, @jni!
by Tony Yu 23 Jul '13

23 Jul '13

Congratulations, Juan. It's great to have you on the team! On Tue, Jul 23, 2013 at 8:00 AM, Juan Nunez-Iglesias <jni.soma(a)gmail.com>wrote: > Woot! Thank you. It's great to be here. I love the project and the team. > > > On Tue, Jul 23, 2013 at 9:29 PM, Stéfan van der Walt <stefan(a)sun.ac.za>wrote: > >> Dear all, >> >> It gives me great pleasure to welcome Juan Nunez-Iglesias to the core >> skimage team! Juan certainly earned his way by suffering through many >> of our most nit picky reviews, and through his excellent >> contributions, including the recent N-d extension to SLIC >> segmentation. >> >> Juan, it's great to have you on board. >> >> Stéfan >> >> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Re: Welcome to the core team, @jni!
by Juan Nunez-Iglesias 23 Jul '13

23 Jul '13

Woot! Thank you. It's great to be here. I love the project and the team. On Tue, Jul 23, 2013 at 9:29 PM, Stéfan van der Walt <stefan(a)sun.ac.za>wrote: > Dear all, > > It gives me great pleasure to welcome Juan Nunez-Iglesias to the core > skimage team! Juan certainly earned his way by suffering through many > of our most nit picky reviews, and through his excellent > contributions, including the recent N-d extension to SLIC > segmentation. > > Juan, it's great to have you on board. > > Stéfan > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Welcome to the core team, @jni!
by Stéfan van der Walt 23 Jul '13

23 Jul '13

Dear all, It gives me great pleasure to welcome Juan Nunez-Iglesias to the core skimage team! Juan certainly earned his way by suffering through many of our most nit picky reviews, and through his excellent contributions, including the recent N-d extension to SLIC segmentation. Juan, it's great to have you on board. Stéfan

1 0

Issue with skimage.transform.warp - 'numpy.float64' object cannot be interpreted as an index
by Alistair Muldal 22 Jul '13

22 Jul '13

Hi all, I've been using skimage.transform.warp to apply affine transformations to a series of movie frames, using this wrapper function: def applytform(img,tmat,**kwargs): """ A wrapper to call skimage.transform.warp for fast similarity transformations. Also returns the mask of the transformed image (a boolean array, True only in pixels containing the transformed image). All keyword arguments are passed directly to warp. """ # warp needs floating point images -1.0 <= I <= 1.0 origtype, imin, imax = img.dtype, img.min(), img.max() tmpimg = (np.float64(img) - imin) / (imax-imin) # tmpimg = np.require(tmpimg,np.float32,'CAOW') # tmat = np.require(tmat,np.float32,'CAOW') # construct a callable transform object, apply it using warp s = SimilarityTransform(matrix=tmat) # import ipdb;ipdb.set_trace() tmpimg = warp(tmpimg,s.inverse,mode='constant',cval=np.nan,**kwargs) mask = ~np.isnan(tmpimg) tmpimg[~mask] = 0. # convert back to original scaling and type img = tmpimg*imax + imin img = img.astype(origtype) return img,mask I've had no problems using skimage v0.7.1, but I've recently tried updating to v0.8.2 and now the same code gives me this error: /home/alistair/src/python/ca_tadpoles2/opt/d/imtransform3.pyc in applytform(img, tmat, **kwar gs) 50 s = SimilarityTransform(matrix=tmat) 51 # import ipdb;ipdb.set_trace() ---> 52 tmpimg = warp(tmpimg,s.inverse,mode='constant',cval=np.nan,**kwargs) 53 mask = ~np.isnan(tmpimg) 54 tmpimg[~mask] = 0. /home/alistair/.venvs/newskimage/lib/python2.7/site-packages/skimage/transform/_geometric.pyc in warp(image, inverse_map, map_args, output_shape, order, mode, cval, reverse_map) 989 dims.append(_warp_fast(image[..., dim], matrix, 990 output_shape=output_shape, --> 991 order=order, mode=mode, cval=cval)) 992 out = np.dstack(dims) 993 if orig_ndim == 2: /home/alistair/.venvs/newskimage/lib/python2.7/site-packages/skimage/transform/_warps_cy.so i n skimage.transform._warps_cy._warp_fast (skimage/transform/_warps_cy.c:1636)() TypeError: 'numpy.float64' object cannot be interpreted as an index I've tried casting the image and the transformation matrix to np.float32, but this gave me a different error: /home/alistair/src/python/ca_tadpoles2/opt/d/imtransform3.pyc in applytform(img, tmat, **kwargs) 50 s = SimilarityTransform(matrix=tmat) 51 # import ipdb;ipdb.set_trace() ---> 52 tmpimg = warp(tmpimg,s.inverse,mode='constant',cval=np.nan,**kwargs) 53 mask = ~np.isnan(tmpimg) 54 tmpimg[~mask] = 0. /home/alistair/.venvs/newskimage/lib/python2.7/site-packages/skimage/transform/_geometric.pyc in warp(image, inverse_map, map_args, output_shape, order, mode, cval, reverse_map) 989 dims.append(_warp_fast(image[..., dim], matrix, 990 output_shape=output_shape, --> 991 order=order, mode=mode, cval=cval)) 992 out = np.dstack(dims) 993 if orig_ndim == 2: /home/alistair/.venvs/newskimage/lib/python2.7/site-packages/skimage/transform/_warps_cy.so in skimage.transform._warps_cy._warp_fast (skimage/transform/_warps_cy.c:1505)() ValueError: Buffer dtype mismatch, expected 'double_t' but got 'float' Any idea what's going on here?

2 3

Re: How to get a region chain code?
by Marianne Corvellec 21 Jul '13

21 Jul '13

Hello, On Tuesday, July 16, 2013 4:46:28 AM UTC-4, Stefan van der Walt wrote: > > This is the third time in about a month that we have the need to determine > geometric properties of regions. The right solution here is to build a > region graph, like I described in the ticket (sorry, I can't check the > number from my phone right now). > I guess this one: https://github.com/scikit-image/scikit-image/issues/622 Marianne

1 0

Re: Cython help: Time taken to execute a function (profiling) changes drastically depending on image size
by Tony Yu 21 Jul '13

21 Jul '13

On Sun, Jul 21, 2013 at 4:45 AM, Chintak Sheth <chintaksheth(a)gmail.com>wrote: > Hi, > > I am writing the code for Image inpainting using Fast Marching Method. > After writing the initial implementation in Python, I am now cythonising > parts of the code [1]. After profiling, the majority of the time is taken > by the 3 function in [2]. In particular, even after cythonizing, > `inpaint_point` takes the majority of the time. > > The results of profiling for 2 different image sizes are here > http://pastebin.com/aTY1bbn6. The first result is for image size of > 100X100 and the second one for 512X512. The mask size which I use is > constant for both cases. So, the number of function calls to > `inpaint_point` for both are the same as can be seem in the profiling > result. > > Still why do I get such a drastic time difference in the 2 cases ? Any > ideas, what I am missing here or doing inefficiently ? > Having different Cython functions using different copies of your data (i.e. one version as uint8, one as float) seems like a bad idea. And making that copy inside of a loop is probably killing performance: https://github.com/chintak/scikit-image/blob/inpaint/skimage/inpaint/_inpai… > Chintak > > [1] https://github.com/chintak/scikit-image/tree/inpaint/skimage/inpaint > > [2] > https://github.com/chintak/scikit-image/blob/inpaint/skimage/inpaint/_inpai… > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Cython help: Time taken to execute a function (profiling) changes drastically depending on image size
by Chintak Sheth 21 Jul '13

21 Jul '13

Hi, I am writing the code for Image inpainting using Fast Marching Method. After writing the initial implementation in Python, I am now cythonising parts of the code [1]. After profiling, the majority of the time is taken by the 3 function in [2]. In particular, even after cythonizing, `inpaint_point` takes the majority of the time. The results of profiling for 2 different image sizes are here http://pastebin.com/aTY1bbn6. The first result is for image size of 100X100 and the second one for 512X512. The mask size which I use is constant for both cases. So, the number of function calls to `inpaint_point` for both are the same as can be seem in the profiling result. Still why do I get such a drastic time difference in the 2 cases ? Any ideas, what I am missing here or doing inefficiently ? Chintak [1] https://github.com/chintak/scikit-image/tree/inpaint/skimage/inpaint [2] https://github.com/chintak/scikit-image/blob/inpaint/skimage/inpaint/_inpai…

1 0

Re: Contiguous Array
by Chintak Sheth 21 Jul '13

21 Jul '13

Hi Ronnie, On Jul 21, 2013 10:00 AM, "Ronnie Ghose" <ronnie.ghose(a)gmail.com> wrote: > > So in skimage/colors why does it matter if the array is contiguous? Is this for Cython operations later? > Yeah it is mainly for using memory views in Cython which is initialized as C contiguous. `cdef some_type[:. ::1] var_name` In thus case ::1 is for C contiguous. Chintak

1 0

Contiguous Array
by Ronnie Ghose 21 Jul '13

21 Jul '13

So in skimage/colors why does it matter if the array is contiguous? Is this for Cython operations later? Thanks, Ronnie

1 0

Re:
by Ronnie Ghose 20 Jul '13

20 Jul '13

Also I was wondering how many you guys were getting - i.e., did I forget to install a component... like i'm going to go add pyamg to the docs next. Actual Error Report: http://pastebin.com/3cSv50YC Thanks, Ronnie On Sat, Jul 20, 2013 at 6:59 PM, Ronnie Ghose <ronnie.ghose(a)gmail.com>wrote: > I'm pretty sure the cython files were not build - but let me try now; I > used bentomaker install before though ... does compile the Cython files for > you? > > > On Sat, Jul 20, 2013 at 6:53 PM, Josh Warner <silvertrumpet999(a)gmail.com>wrote: > >> Correction, that should be `python setup.py build_ext -i`, run at package >> root. >> >> >> On Saturday, July 20, 2013 5:52:21 PM UTC-5, Josh Warner wrote: >>> >>> Run again with `make test` and copy the entire error report, please. >>> That section at the end of coverage just tells us there were 11 errors, not >>> where they came from or why. >>> >>> However, because so few tests were ran I'll go out on a limb and guess >>> the Cython files are not compiled. >>> >>> Try `python build_ext -i` and then running `make test` again; if I'm >>> right that would fix it. >>> >>> Regards, >>> Josh >>> >>> On Friday, July 19, 2013 10:53:34 PM UTC-5, Ronnie Ghose wrote: >>>> >>>> how many errors are you getting from `make coverage` ? at the moment >>>> i'm getting: >>>> >>>> . . . >>>> . . . >>>> skimage.viewer.viewers.core 179 142 21% 26-29, >>>> 74-120, 124-139, 142-153, 157-161, 164-166, 176-194, 197, 200-205, 212-214, >>>> 217, 221, 225-239, 242, 246-247, 251, 254-257, 261-266, 298-310, 319-330, >>>> 338, 341-351 >>>> skimage.viewer.widgets 2 0 100% >>>> skimage.viewer.widgets.core 130 103 21% 33-37, >>>> 41-42, 45, 51-59, 63, 67, 99-165, 169-171, 175-186, 189, 192, 196-199, >>>> 203-205, 228-241, 247, 251, 255 >>>> skimage.viewer.widgets.history 70 52 26% 24-39, >>>> 42-44, 48, 55-72, 75-82, 85-92, 96-104 >>>> ------------------------------**------------------------------** >>>> --------- >>>> TOTAL 3741 1338 64% >>>> ------------------------------**------------------------------** >>>> ---------- >>>> Ran 243 tests in 4.576s >>>> >>>> FAILED (SKIP=17, errors=11) >>>> >>>> >>>> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > >

1 0

Re:
by Josh Warner 20 Jul '13

20 Jul '13

Run again with `make test` and copy the entire error report, please. That section at the end of coverage just tells us there were 11 errors, not where they came from or why. However, because so few tests were ran I'll go out on a limb and guess the Cython files are not compiled. Try `python build_ext -i` and then running `make test` again; if I'm right that would fix it. Regards, Josh On Friday, July 19, 2013 10:53:34 PM UTC-5, Ronnie Ghose wrote: > > how many errors are you getting from `make coverage` ? at the moment i'm > getting: > > . . . > . . . > skimage.viewer.viewers.core 179 142 21% 26-29, > 74-120, 124-139, 142-153, 157-161, 164-166, 176-194, 197, 200-205, 212-214, > 217, 221, 225-239, 242, 246-247, 251, 254-257, 261-266, 298-310, 319-330, > 338, 341-351 > skimage.viewer.widgets 2 0 100% > skimage.viewer.widgets.core 130 103 21% 33-37, > 41-42, 45, 51-59, 63, 67, 99-165, 169-171, 175-186, 189, 192, 196-199, > 203-205, 228-241, 247, 251, 255 > skimage.viewer.widgets.history 70 52 26% 24-39, > 42-44, 48, 55-72, 75-82, 85-92, 96-104 > --------------------------------------------------------------------- > TOTAL 3741 1338 64% > ---------------------------------------------------------------------- > Ran 243 tests in 4.576s > > FAILED (SKIP=17, errors=11) > > >

2 2

(no subject)
by Ronnie Ghose 20 Jul '13

20 Jul '13

how many errors are you getting from `make coverage` ? at the moment i'm getting: . . . . . . skimage.viewer.viewers.core 179 142 21% 26-29, 74-120, 124-139, 142-153, 157-161, 164-166, 176-194, 197, 200-205, 212-214, 217, 221, 225-239, 242, 246-247, 251, 254-257, 261-266, 298-310, 319-330, 338, 341-351 skimage.viewer.widgets 2 0 100% skimage.viewer.widgets.core 130 103 21% 33-37, 41-42, 45, 51-59, 63, 67, 99-165, 169-171, 175-186, 189, 192, 196-199, 203-205, 228-241, 247, 251, 255 skimage.viewer.widgets.history 70 52 26% 24-39, 42-44, 48, 55-72, 75-82, 85-92, 96-104 --------------------------------------------------------------------- TOTAL 3741 1338 64% ---------------------------------------------------------------------- Ran 243 tests in 4.576s FAILED (SKIP=17, errors=11)

1 0

Re: color difference functions
by Matt Terry 19 Jul '13

19 Jul '13

Sounds good. I'll send a PR when I get the tests finished. -matt On Friday, July 19, 2013 2:03:58 AM UTC-7, Stefan van der Walt wrote: > > Hi Matt > > On Fri, Jul 19, 2013 at 1:36 AM, Matt Terry <matt....(a)gmail.com<javascript:>> > wrote: > > Would color difference functions [1] be of interest to skimage? I have > the > > distance functions for CIE76, CIEDE94, CIEDE2000 and CMC l:c if you > think > > the would be of general interest. > > Those fit in very well with the scope of the ``skimage.color`` module. > I think especially if we can have docstrings that teach the user a > bit more about the topic, this would be highly useful in the classroom > too. > > Stéfan >

1 0

color difference functions
by Matt Terry 19 Jul '13

19 Jul '13

Hi, Would color difference functions [1] be of interest to skimage? I have the distance functions for CIE76, CIEDE94, CIEDE2000 and CMC l:c if you think the would be of general interest. -matt [1] http://en.wikipedia.org/wiki/Color_difference

2 1

Re: IRC Channel
by Stéfan van der Walt 19 Jul '13

19 Jul '13

On Fri, Jul 19, 2013 at 3:48 AM, Ronnie Ghose <ronnie.ghose(a)gmail.com> wrote: > is there no IRC channel for scikit-image? if not - why? There is #skimage on freenode.net, but not a lot of activity. Stéfan

1 0

Re: IRC Channel
by Johannes Schönberger 19 Jul '13

19 Jul '13

I can only speak for myself, but I personally do not use IRC… Johannes Schönberger Am 19.07.2013 um 03:48 schrieb Ronnie Ghose <ronnie.ghose(a)gmail.com>: > is there no IRC channel for scikit-image? if not - why? > > -- > You received this message because you are subscribed to the Google Groups "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > >

1 0

IRC Channel
by Ronnie Ghose 19 Jul '13

19 Jul '13

is there no IRC channel for scikit-image? if not - why?

1 0

Re: Failed to import LineModel, ransac
by Michael Aye 18 Jul '13

18 Jul '13

> > > I didn't think EPD offered dev versions of packages. I can only see 0.8.2. > In any case, it depends on how recently the dev version was updated. The > RANSAC code was only merged into master about 2 months ago. (Any pulls from > master in the last 4 months will be labeled 0.9dev.) > Yeah, sorry for that. Must have pulled master in at sometime for some interesting feature. I am still regularly confused by the fact that enpkg --remove does leave the egg in site-packages, and for some reason I don't have the git folder of skimage anymore, so I thought it must have come via enpkg.

1 0

ImportError: cannot import name LineModel
by Fabian Peña Arellano 18 Jul '13

18 Jul '13

Dear all, I am trying to run the ransac model estimation example given on the documentation. However, Python cannot load the line model: Traceback (most recent call last): File "plot_ransac.py", line 13, in <module> from skimage.measure import LineModel, ransac ImportError: cannot import name LineModel I am using the latest stable version 0.8.2 in Ubuntu. Are the LineModel and ransac routines available on the development version only? Thank you very much in advance. Regards, Fabián

2 1

Re: Failed to import LineModel, ransac
by Johannes Schönberger 18 Jul '13

18 Jul '13

@all You must download the latest development version on github to use the Line, Circle and EllipseModel and ransac. They were added just about 1 month ago. 0.9 is the dev version. 0.8 the stable version. Johannes Schönberger Am 18.07.2013 um 20:42 schrieb "Jean-Patrick Pommier" <jeanpatrick.pommier(a)gmail.com<mailto:jeanpatrick.pommier@gmail.com>>: The example using LineModel <http://scikit-image.org/docs/dev/auto_examples/plot_ransac.html> mentions skimage 0.9. To use LineModel, should I try to use the stable version ? 2013/7/18 Stéfan van der Walt <stefan(a)sun.ac.za<mailto:stefan@sun.ac.za>> We should update the online docs to point to the last stable version instead of to dev (as has been suggested by others in the past)--I see it causes some confusion. Stéfan On Thu, Jul 18, 2013 at 12:42 PM, Jean-Patrick Pommier <jeanpatrick.pommier(a)gmail.com<mailto:jeanpatrick.pommier@gmail.com>> wrote: > I forget to mention that skimage is running on ubuntu 13.04 and was > installed with easy_install. > > > 2013/7/18 Tony Yu <tsyu80(a)gmail.com<mailto:tsyu80@gmail.com>> >> >> >> >> >> On Wed, Jul 17, 2013 at 6:43 PM, Michael Aye <kmichael.aye(a)gmail.com<mailto:kmichael.aye@gmail.com>> >> wrote: >>> >>> I also have 0.9dev via EPD, and also do not have LineModel or ransac >>> inside the measure module. >> >> >> I didn't think EPD offered dev versions of packages. I can only see 0.8.2. >> In any case, it depends on how recently the dev version was updated. The >> RANSAC code was only merged into master about 2 months ago. (Any pulls from >> master in the last 4 months will be labeled 0.9dev.) >> >> Best, >> -Tony >> >>> On Wednesday, July 17, 2013 10:34:01 AM UTC-7, Jean-Patrick Pommier >>> wrote: >>>> >>>> Hi, >>>> >>>> >>>> When doing from ipython: >>>> >>>> print skimage.__version__ >>>> from skimage.measure import LineModel >>>> >>>> I get: >>>> >>>> >>>> --------------------------------------------------------------------------- >>>> ImportError Traceback (most recent call >>>> last) >>>> <ipython-input-55-c6b2bd2de1ce> in <module>() >>>> 1 print skimage.__version__ >>>> ----> 2 from skimage.measure import LineModel >>>> >>>> ImportError: cannot import name LineModel >>>> >>>> 0.9dev >>>> >>>> >>>> Is it a version problem ? >>>> >>>> Jean-Patrick >>>> >>>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "scikit-image" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to scikit-image+unsubscribe(a)googlegroups.com<mailto:scikit-image%2Bunsubscribe@googlegroups.com>. >>> >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >>> >> >> >> -- >> You received this message because you are subscribed to a topic in the >> Google Groups "scikit-image" group. >> To unsubscribe from this topic, visit >> https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. >> To unsubscribe from this group and all of its topics, send an email to >> scikit-image+unsubscribe(a)googlegroups.com<mailto:scikit-image%2Bunsubscribe@googlegroups.com>. >> >> For more options, visit https://groups.google.com/groups/opt_out. >> >> > > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com<mailto:scikit-image%2Bunsubscribe@googlegroups.com>. > For more options, visit https://groups.google.com/groups/opt_out. > > -- You received this message because you are subscribed to a topic in the Google Groups "scikit-image" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. To unsubscribe from this group and all of its topics, send an email to scikit-image+unsubscribe(a)googlegroups.com<mailto:scikit-image%2Bunsubscribe@googlegroups.com>. For more options, visit https://groups.google.com/groups/opt_out. -- You received this message because you are subscribed to the Google Groups "scikit-image" group. To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image+unsubscribe(a)googlegroups.com<mailto:scikit-image+unsubscribe@googlegroups.com>. For more options, visit https://groups.google.com/groups/opt_out.

2 1

Re: Failed to import LineModel, ransac
by Jean-Patrick Pommier 18 Jul '13

18 Jul '13

The example using LineModel <http://scikit-image.org/docs/dev/auto_examples/plot_ransac.html>mentions skimage 0.9. To use *LineModel*, should I try to use the stable version ? 2013/7/18 Stéfan van der Walt <stefan(a)sun.ac.za> > We should update the online docs to point to the last stable version > instead of to dev (as has been suggested by others in the past)--I see > it causes some confusion. > > Stéfan > > On Thu, Jul 18, 2013 at 12:42 PM, Jean-Patrick Pommier > <jeanpatrick.pommier(a)gmail.com> wrote: > > I forget to mention that skimage is running on ubuntu 13.04 and was > > installed with easy_install. > > > > > > 2013/7/18 Tony Yu <tsyu80(a)gmail.com> > >> > >> > >> > >> > >> On Wed, Jul 17, 2013 at 6:43 PM, Michael Aye <kmichael.aye(a)gmail.com> > >> wrote: > >>> > >>> I also have 0.9dev via EPD, and also do not have LineModel or ransac > >>> inside the measure module. > >> > >> > >> I didn't think EPD offered dev versions of packages. I can only see > 0.8.2. > >> In any case, it depends on how recently the dev version was updated. The > >> RANSAC code was only merged into master about 2 months ago. (Any pulls > from > >> master in the last 4 months will be labeled 0.9dev.) > >> > >> Best, > >> -Tony > >> > >>> On Wednesday, July 17, 2013 10:34:01 AM UTC-7, Jean-Patrick Pommier > >>> wrote: > >>>> > >>>> Hi, > >>>> > >>>> > >>>> When doing from ipython: > >>>> > >>>> print skimage.__version__ > >>>> from skimage.measure import LineModel > >>>> > >>>> I get: > >>>> > >>>> > >>>> > --------------------------------------------------------------------------- > >>>> ImportError Traceback (most recent call > >>>> last) > >>>> <ipython-input-55-c6b2bd2de1ce> in <module>() > >>>> 1 print skimage.__version__ > >>>> ----> 2 from skimage.measure import LineModel > >>>> > >>>> ImportError: cannot import name LineModel > >>>> > >>>> 0.9dev > >>>> > >>>> > >>>> Is it a version problem ? > >>>> > >>>> Jean-Patrick > >>>> > >>>> > >>> -- > >>> You received this message because you are subscribed to the Google > Groups > >>> "scikit-image" group. > >>> To unsubscribe from this group and stop receiving emails from it, send > an > >>> email to scikit-image+unsubscribe(a)googlegroups.com. > >>> > >>> For more options, visit https://groups.google.com/groups/opt_out. > >>> > >>> > >> > >> > >> -- > >> You received this message because you are subscribed to a topic in the > >> Google Groups "scikit-image" group. > >> To unsubscribe from this topic, visit > >> https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. > >> To unsubscribe from this group and all of its topics, send an email to > >> scikit-image+unsubscribe(a)googlegroups.com. > >> > >> For more options, visit https://groups.google.com/groups/opt_out. > >> > >> > > > > > > -- > > You received this message because you are subscribed to the Google Groups > > "scikit-image" group. > > To unsubscribe from this group and stop receiving emails from it, send an > > email to scikit-image+unsubscribe(a)googlegroups.com. > > For more options, visit https://groups.google.com/groups/opt_out. > > > > > > -- > You received this message because you are subscribed to a topic in the > Google Groups "scikit-image" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. > To unsubscribe from this group and all of its topics, send an email to > scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Re: GLCM with scikit-image
by Neil 18 Jul '13

18 Jul '13

I wouldn't worry too much about the warning. Whenever you convert an image from floating point (64 bits per pixel) to bytes (8 bits per pixel), there may be a loss of precision. The lines: im = skimage.img_as_ubyte(im) im /= 32 are just a quick way to scale the pixel intensities to have 8 different levels. The first step converts the image from floats to bytes (this is where you get the warning). We now have an image with 256 levels (from 0 to 255). After dividing by 32, there are only 8 possible levels (from 0 to 7). It is likely that matlab use a different procedure for converting from a floating point image to an image with only 8 levels. The [0][0] is to specify which offset you want. The first 0 gives the index for the distances, and the second is the index for the angles. In this example, we are only computing one offset (distance=1 and angle=0). However, in many real-life applications you will want to compute the statistics for a number of offsets to capture differences in scale and orientation. Neil On Monday, 8 July 2013 14:00:11 UTC+1, ely...(a)mail.com wrote: > > Hi, > > Thanks a lot for the reply. > > Indeed graycoprops normalizes the gray-level co-occurrence matrix: > > “graycoprops normalizes the gray-level co-occurrence matrix (GLCM) so that > the sum of its elements is equal to 1. Each element (r,c) in the normalized > GLCM is the joint probability occurrence of pixel pairs with a defined > spatial relationship having gray level values r and c in the image. > graycoprops uses the normalized GLCM to calculate properties.” ( > http://www.mathworks.co.uk/help/images/ref/graycoprops.html) > > The modified script: > import skimage > > from skimage.io import imread > from skimage.feature import greycomatrix > from skimage.feature import greycoprops > > im = imread('C:/Users/Asher_dell/Desktop/ImgTemp/python.jpg', as_grey=True) > > im = skimage.img_as_ubyte(im) > im /= 32 > g = skimage.feature.greycomatrix(im, [1], [0], levels=8, symmetric=False, > normed=True) > > contrast= skimage.feature.greycoprops(g, 'contrast')[0][0] > energy= skimage.feature.greycoprops(g, 'energy')[0][0] > homogeneity= skimage.feature.greycoprops(g, 'homogeneity')[0][0] > correlation=skimage.feature.greycoprops(g, 'correlation')[0][0] > dissimilarity=skimage.feature.greycoprops(g, 'dissimilarity')[0][0] > ASM=skimage.feature.greycoprops(g, 'ASM')[0][0] > > > print('contrast is: ', contrast) > print('energy is: ', energy) > print('homogeneity is: ', homogeneity) > print('correlation is: ', correlation) > print('dissimilarity is: ', dissimilarity) > print('ASM is: ', ASM) > > > Output: > skimage.dtype_converter: WARNING: Possible precision loss when converting > from float64 to uint8 > contrast is: 0.301542207792 > energy is: 0.29069020973 > homogeneity is: 0.883463991917 > correlation is: 0.971624675221 > dissimilarity is: 0.243464091878 > ASM is: 0.0845007980331 > > Based on the output I have few more questions on the modified script: > - Does the "skimage.dtype_converter: WARNING: Possible precision loss when > converting from float64 to uint8" means that the output values are wrong? > - Can you please explain me why you use these lines: > > im = skimage.img_as_ubyte(im) > im /= 32 > - Why do you use the [0][0] when you call skimage.feature.greycoprops > (e.g.,skimage.feature.greycoprops(g, 'contrast')[0][0])? > > Thanks a lot again.

4 4

Re: Failed to import LineModel, ransac
by Stéfan van der Walt 18 Jul '13

18 Jul '13

We should update the online docs to point to the last stable version instead of to dev (as has been suggested by others in the past)--I see it causes some confusion. Stéfan On Thu, Jul 18, 2013 at 12:42 PM, Jean-Patrick Pommier <jeanpatrick.pommier(a)gmail.com> wrote: > I forget to mention that skimage is running on ubuntu 13.04 and was > installed with easy_install. > > > 2013/7/18 Tony Yu <tsyu80(a)gmail.com> >> >> >> >> >> On Wed, Jul 17, 2013 at 6:43 PM, Michael Aye <kmichael.aye(a)gmail.com> >> wrote: >>> >>> I also have 0.9dev via EPD, and also do not have LineModel or ransac >>> inside the measure module. >> >> >> I didn't think EPD offered dev versions of packages. I can only see 0.8.2. >> In any case, it depends on how recently the dev version was updated. The >> RANSAC code was only merged into master about 2 months ago. (Any pulls from >> master in the last 4 months will be labeled 0.9dev.) >> >> Best, >> -Tony >> >>> On Wednesday, July 17, 2013 10:34:01 AM UTC-7, Jean-Patrick Pommier >>> wrote: >>>> >>>> Hi, >>>> >>>> >>>> When doing from ipython: >>>> >>>> print skimage.__version__ >>>> from skimage.measure import LineModel >>>> >>>> I get: >>>> >>>> >>>> --------------------------------------------------------------------------- >>>> ImportError Traceback (most recent call >>>> last) >>>> <ipython-input-55-c6b2bd2de1ce> in <module>() >>>> 1 print skimage.__version__ >>>> ----> 2 from skimage.measure import LineModel >>>> >>>> ImportError: cannot import name LineModel >>>> >>>> 0.9dev >>>> >>>> >>>> Is it a version problem ? >>>> >>>> Jean-Patrick >>>> >>>> >>> -- >>> You received this message because you are subscribed to the Google Groups >>> "scikit-image" group. >>> To unsubscribe from this group and stop receiving emails from it, send an >>> email to scikit-image+unsubscribe(a)googlegroups.com. >>> >>> For more options, visit https://groups.google.com/groups/opt_out. >>> >>> >> >> >> -- >> You received this message because you are subscribed to a topic in the >> Google Groups "scikit-image" group. >> To unsubscribe from this topic, visit >> https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. >> To unsubscribe from this group and all of its topics, send an email to >> scikit-image+unsubscribe(a)googlegroups.com. >> >> For more options, visit https://groups.google.com/groups/opt_out. >> >> > > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > >

1 0

seeking advice on HoG applicability
by Lisa Torrey 18 Jul '13

18 Jul '13

Hi folks - I'm looking at an image classification problem, and wondering whether HoG should be applicable to it. If any of you are willing to take a look and give some advice, that would be wonderful. I have a machine learning background, but working with image data is a new area for me. The problem is to distinguish between two types of moss. Type 1 tends to consist of upright stalks with few or no branches. Type 2 tends to have secondary branches coming off the primary stalk. There's quite a bit of visual diversity within these types. I've linked some images below. Type 1: http://myslu.stlawu.edu/~ltorrey/moss/andrea_rothii.jpg http://myslu.stlawu.edu/~ltorrey/moss/mnium_spinulosum.jpg Type 2: http://myslu.stlawu.edu/~ltorrey/moss/climacium_americanum.jpg http://myslu.stlawu.edu/~ltorrey/moss/rhytidiadelphus_triquetrus.jpg When I came across the Dalal paper, I thought my problem might have something in common with the pedestrian detection problem, so I tried extracting HoG features and feeding them into an SVM classifier. This failed miserably - the SVM does no better than random guessing. I'm now trying to weigh potential reasons. The first possible reason on my list is the diversity among mosses of the same type. There isn't necessarily a "type 1 shape" and a "type 2 shape," at least not to the degree that there's a "pedestrian shape." Perhaps this means HoG isn't really the right approach to my problem after all? Other reasons may include: - I have much less data. (Just 77 positives and 78 negatives, compared to Dalal's 1239 and 12180.) - My images aren't all the same size, like the pedestrian images are. (I'm not sure if this would matter?) - My images are much higher resolution. (I've been downscaling them by a factor of 8, but the feature vectors are still enormous.) - I'm just using default parameters so far. (In the absence of any signal, tweaking seems unproductive.) Any thoughts or suggestions would be welcome! -Lisa

4 6

Re: Failed to import LineModel, ransac
by Jean-Patrick Pommier 18 Jul '13

18 Jul '13

I forget to mention that skimage is running on ubuntu 13.04 and was installed with easy_install. 2013/7/18 Tony Yu <tsyu80(a)gmail.com> > > > > On Wed, Jul 17, 2013 at 6:43 PM, Michael Aye <kmichael.aye(a)gmail.com>wrote: > >> I also have 0.9dev via EPD, and also do not have LineModel or ransac >> inside the measure module. > > > I didn't think EPD offered dev versions of packages. I can only see 0.8.2. > In any case, it depends on how recently the dev version was updated. The > RANSAC code was only merged into master about 2 months ago. (Any pulls from > master in the last 4 months will be labeled 0.9dev.) > > Best, > -Tony > > On Wednesday, July 17, 2013 10:34:01 AM UTC-7, Jean-Patrick Pommier wrote: >>> >>> Hi, >>> >>> >>> When doing from ipython: >>> >>> print skimage.__version__ >>> from skimage.measure import LineModel >>> >>> I get: >>> >>> ------------------------------**------------------------------**---------------ImportError Traceback (most recent call last)<ipython-input-55-**c6b2bd2de1ce> in <module>() 1 print skimage.__version__----> 2 from skimage.measure import LineModel >>> ImportError: cannot import name LineModel >>> >>> >>> 0.9dev >>> >>> >>> Is it a version problem ? >>> >>> Jean-Patrick >>> >>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > > -- > You received this message because you are subscribed to a topic in the > Google Groups "scikit-image" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/scikit-image/24fxSmErl1c/unsubscribe. > To unsubscribe from this group and all of its topics, send an email to > scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0

Failed to import LineModel, ransac
by Jean-Patrick Pommier 18 Jul '13

18 Jul '13

Hi, When doing from ipython: print skimage.__version__ from skimage.measure import LineModel I get: ---------------------------------------------------------------------------ImportError Traceback (most recent call last)<ipython-input-55-c6b2bd2de1ce> in <module>() 1 print skimage.__version__----> 2 from skimage.measure import LineModel ImportError: cannot import name LineModel 0.9dev Is it a version problem ? Jean-Patrick

3 2

Re: How to get a region chain code?
by Valeriy Sokolov 16 Jul '13

16 Jul '13

I would like to help with the implementation. I am going to make draft which will work for my current case. But I will need some guidance and review to add it to the skimage code base. It would be helpful if you have some approach in mind... I am going to get shape boundary pixels (with morphology) and then apply something like pixel grid graph traverse of the boundary with Cython. On Tue, Jul 16, 2013 at 12:46 PM, Stéfan van der Walt <stefan(a)sun.ac.za>wrote: > This is the third time in about a month that we have the need to determine > geometric properties of regions. The right solution here is to build a > region graph, like I described in the ticket (sorry, I can't check the > number from my phone right now). Would anyone be interested in helping to > implement that? > > In the meantime we can implement the border finding manually, and later > just slot in the new structure. > > Stéfan > On 16 Jul 2013 07:03, "Chintak Sheth" <chintaksheth(a)gmail.com> wrote: > >> Hi Valeriy, >> >> Wow a compression technique! That would surely be a great addition. >> >> > I am studying a shape descriptors and would like to compute a chain >> code description of a labeled shape. I was trying to look into docs but >> have not find a direct way to get this region property with skimage. Is >> there any way to do it in a few calls which I missed? >> > >> >> - Extract the boundary. `skimage.morphology.erosion` using >> `skimage.morphology.disk` structuring element with radius 1. Then subtract >> this from the original image. >> >> - Label the boundaries. If you already have a labeled image, then that >> saves you a function call else there is `label` function in >> `skimage.morphology`. >> >> - Use `np.transpose(np.where(label == 1))` to generate the indices of >> pixels marked as label 1. This will return a `list` of indices. You can >> simply use the first as the starting point. >> >> This should help set up the stage for the traversal. Hope this helps a >> bit. >> >> Chintak. >> >> -- >> You received this message because you are subscribed to the Google Groups >> "scikit-image" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to scikit-image+unsubscribe(a)googlegroups.com. >> For more options, visit https://groups.google.com/groups/opt_out. >> >> >> > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > > -- Best regards, Valeriy Sokolov.

1 0

Re: How to get a region chain code?
by Valeriy Sokolov 16 Jul '13

16 Jul '13

On Tue, Jul 16, 2013 at 9:03 AM, Chintak Sheth <chintaksheth(a)gmail.com>wrote: > Hi Valeriy, > > Wow a compression technique! That would surely be a great addition. > Compression is cool no doubt, but I need it to turn the labeled shape into a polygon =) > > I am studying a shape descriptors and would like to compute a chain code > description of a labeled shape. I was trying to look into docs but have not > find a direct way to get this region property with skimage. Is there any > way to do it in a few calls which I missed? > > > > - Extract the boundary. `skimage.morphology.erosion` using > `skimage.morphology.disk` structuring element with radius 1. Then subtract > this from the original image. > I suppose `skimage.morphology.dilation` would work better. It will work in case we computing chain code for the line of 1-pixel width. > - Label the boundaries. If you already have a labeled image, then that > saves you a function call else there is `label` function in > `skimage.morphology`. > > - Use `np.transpose(np.where(label == 1))` to generate the indices of > pixels marked as label 1. This will return a `list` of indices. You can > simply use the first as the starting point. > > This should help set up the stage for the traversal. Hope this helps a bit. > The traverse is the most interesting part in this question =) The other things do not introduce order on the pixels and could be made with current skimage easily. > Chintak. > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > > -- Best regards, Valeriy Sokolov.

1 0

Re: How to get a region chain code?
by Stéfan van der Walt 16 Jul '13

16 Jul '13

This is the third time in about a month that we have the need to determine geometric properties of regions. The right solution here is to build a region graph, like I described in the ticket (sorry, I can't check the number from my phone right now). Would anyone be interested in helping to implement that? In the meantime we can implement the border finding manually, and later just slot in the new structure. Stéfan On 16 Jul 2013 07:03, "Chintak Sheth" <chintaksheth(a)gmail.com> wrote: > Hi Valeriy, > > Wow a compression technique! That would surely be a great addition. > > > I am studying a shape descriptors and would like to compute a chain code > description of a labeled shape. I was trying to look into docs but have not > find a direct way to get this region property with skimage. Is there any > way to do it in a few calls which I missed? > > > > - Extract the boundary. `skimage.morphology.erosion` using > `skimage.morphology.disk` structuring element with radius 1. Then subtract > this from the original image. > > - Label the boundaries. If you already have a labeled image, then that > saves you a function call else there is `label` function in > `skimage.morphology`. > > - Use `np.transpose(np.where(label == 1))` to generate the indices of > pixels marked as label 1. This will return a `list` of indices. You can > simply use the first as the starting point. > > This should help set up the stage for the traversal. Hope this helps a bit. > > Chintak. > > -- > You received this message because you are subscribed to the Google Groups > "scikit-image" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to scikit-image+unsubscribe(a)googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > > >

1 0