[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [pygame] BUG: JPEG colors saving a pygame.camera surface



Err, oops, I should have proofread that.  It should state:

camlist = camera.list_cameras()
if len(camlist) > 0:
 cam = camera.Camera(camlist[0])

Just a caveat though.  The way list_cameras is currently set up, it
will also detect v4l cameras, which are not actually supported.  I
will most likely remove this feature before the 1.9 release.

Nirav

On Mon, Apr 20, 2009 at 11:04 AM, Nirav Patel <olpc@xxxxxxxxxxxxxx> wrote:
> I'm not sure what it would be without seeing the error output, but
> there is a way around that.  camera.list_cameras() will return a list
> of the cameras plugged into the computer. If the list is empty, you
> know it found no cameras.  You could use it as follows:
>
> camlist = camera.list_cameras()
> if len(cameras) > 0:
>  cam = camera.Camera(camlist[0])
>
> Nirav
>
> On Mon, Apr 20, 2009 at 9:46 AM, Alexandre Quessy <alexandre@xxxxxxxxxx> wrote:
>> Hello René,
>> We save frames in lists. You can take a frame either by pressing space
>> bar, or using an intervalometer. (for timelapses) The right region of
>> the display shows the resulting animation in a loop. (for now) The
>> left region shows the current video input, plus an optional onion skin
>> of the last recorded frame. A screen shot will be online soon.
>>
>> By the way, there is an other bug. With V4L2, if there is no camera
>> plugged-in, sometimes, I get a null pointer error and it says "Pygame
>> parachute", or something. And then it crashes. I have heard it's v4l2
>> that is a bit flaky, but that some higher-level wrapper, such as
>> Gstreamer, make sure things don't crash when using v4l2. Of course,
>> bugs are not reproducable when you need them.... (so that I could
>> copy-paste the output here)
>>
>> a
>>
>>
>>
>> 2009/4/19 René Dudfield <renesd@xxxxxxxxx>:
>>> looks pretty cool :)
>>>
>>> How does it work?
>>>
>>> I imagine you can take snapshots, and then do playbacks?  Or can you do a
>>> video recording too?  So in video mode, it'd take snapshots at a given frame
>>> rate.
>>>
>>> Are there any screen shots?
>>>
>>>
>>>
>>>
>>> On Sat, Apr 18, 2009 at 12:54 PM, Alexandre Quessy <alexandre@xxxxxxxxxx>
>>> wrote:
>>>>
>>>> Problem solved !!
>>>> Thanks a lot Nirav.
>>>>
>>>> If you want to try toonloop, you can checkout the code and run it.
>>>> The installation instructions for Ubuntu are in here :
>>>> http://code.google.com/p/toonloop/source/browse/trunk/doc/INSTALL.txt
>>>>
>>>> Try ./toonloop.py --help for startup flags, and press "h" while it is
>>>> running for instructions.
>>>> It needs a valid pygame camera, of course.
>>>> See http://toonloop.com for more informations.
>>>> We need beta testers too. And I think this is a wonderful application
>>>> of pygame.camera.
>>>>
>>>> a
>>>>
>>>> 2009/4/17 Nirav Patel <olpc@xxxxxxxxxxxxxx>:
>>>> > Also, forgot to mention this, but thanks for using the camera module!
>>>> > It has yet to experience widespread use, so this kind of beta testing
>>>> > is great to get it ready for the next pygame release.
>>>> >
>>>> > Nirav
>>>> >
>>>> > On Fri, Apr 17, 2009 at 11:17 AM, Nirav Patel <olpc@xxxxxxxxxxxxxx>
>>>> > wrote:
>>>> >> Ah yes, this seems to be a bug in both JPEG saving and in the camera
>>>> >> module.  Fun stuff.
>>>> >>
>>>> >> The JPEG assumes that any 24 bit surface is going to be in the exact
>>>> >> color order that it expects.  Which is fine, except that the default
>>>> >> color order for a 24 bit Surface is not that.
>>>> >>
>>>> >> This also led me to find a bug in the camera module though, where if
>>>> >> you are using a 24 bit surface, it will assume you want the default
>>>> >> color order, regardless of if you hand it a Surface to use that has a
>>>> >> different color order.
>>>> >>
>>>> >> I commited the one line fix for JPEG, which just does an additional
>>>> >> check to see if it needs to create a new surface or not.  Please test
>>>> >> it if you can.  The fix for the camera module may take longer, as I
>>>> >> need to rethink some assumptions, though in the vast majority of use
>>>> >> cases, users should not be effected.
>>>> >>
>>>> >> Nirav
>>>> >>
>>>> >> On Fri, Apr 17, 2009 at 9:48 AM, René Dudfield <renesd@xxxxxxxxx>
>>>> >> wrote:
>>>> >>> Hi,
>>>> >>>
>>>> >>> can you please let me know the result of:
>>>> >>>
>>>> >>> cam_surf = X
>>>> >>> normal_surf = pygame.Surface((1,1))
>>>> >>> surfs = cam_surf, normal_surf
>>>> >>>
>>>> >>> for s in surfs:
>>>> >>>     print s.get_losses(), s.get_masks(), s.get_shifts()
>>>> >>>
>>>> >>> I think maybe the jpeg saving code isn't respecting one of those.
>>>> >>> Actually... don't worry, I'm pretty sure that's the cause of it.
>>>> >>> Will fix
>>>> >>> soon.
>>>> >>>
>>>> >>> cheers,
>>>> >>>
>>>> >>>
>>>> >>>
>>>> >>> On Fri, Apr 17, 2009 at 11:14 PM, Alexandre Quessy
>>>> >>> <alexandre@xxxxxxxxxx>
>>>> >>> wrote:
>>>> >>>>
>>>> >>>> Hello Pygame people,
>>>> >>>> This is my first post on this list, and more might follow since I am
>>>> >>>> using pygame for ToonLoop, a stop motion software. The new
>>>> >>>> pygame.camera module works for me and this is very good job. Thanks
>>>> >>>> for contributing that !
>>>> >>>>
>>>> >>>> I think I found a bug ! Hopefully it is only my code that is wrong
>>>> >>>> and
>>>> >>>> this is easy to fix.
>>>> >>>>
>>>> >>>> When I save a Pygame surface that I obtained using pygame.camera to a
>>>> >>>> JPEG using pygame.image.save, the colors are messed up. It looks like
>>>> >>>> the red and blue channels are interchanged. Thus, maybe my surface is
>>>> >>>> RBG, whereas pygame.image expects RGB. A camera image doesn't contain
>>>> >>>> any alpha channel usually.
>>>> >>>>
>>>> >>>> When I display the surface as a pygame sprite the colors are OK.
>>>> >>>> When I display the surface as an OpenGL texture the colors are OK.
>>>> >>>> (using tostring(surface, "RGBX", True))
>>>> >>>> When I save the surface as an other format such as PNG or BMP the
>>>> >>>> colors
>>>> >>>> are OK.
>>>> >>>> When I use a surface obtained by loading a JPG image, the colors are
>>>> >>>> OK.
>>>> >>>> The bug only occurs when I save a surface obtained using the
>>>> >>>> pygame.camera module.
>>>> >>>> It consistently happened on 3 Linux computers.
>>>> >>>>
>>>> >>>> I use Pygame compiled from today's SVN with Python 2.5.2 on Ubuntu
>>>> >>>> GNU/Linux 8.10 using a V4L2 device. (a WinTV card)
>>>> >>>>
>>>> >>>> A short code snippet to reproduce the bug:
>>>> >>>> http://rafb.net/p/gccaJG37.html
>>>> >>>>
>>>> >>>> A JPEG to see how the output looks like :
>>>> >>>> http://toonloop.com/static/tmp/image_color_test_out.jpg
>>>> >>>> A correct image in an other format to compare :
>>>> >>>> http://toonloop.com/static/tmp/image_color_test_out.png
>>>> >>>>
>>>> >>>> If you want to download the code and the test image I use :
>>>> >>>> http://toonloop.com/static/tmp/bug_color_jpeg.tar.gz
>>>> >>>> I also convert the colorbars.jpg files to surface and back to a JPG
>>>> >>>> file for comparison. It works flawlessly.
>>>> >>>>
>>>> >>>> --
>>>> >>>> Alexandre Quessy
>>>> >>>> http://alexandre.quessy.net/
>>>> >>>
>>>> >>>
>>>> >>
>>>> >
>>>>
>>>>
>>>>
>>>> --
>>>> Alexandre Quessy
>>>> http://alexandre.quessy.net/
>>>
>>>
>>
>>
>>
>> --
>> Alexandre Quessy
>> http://alexandre.quessy.net/
>>
>