video

Little has been spoken of the proper use of the OpenCL convertor, which is hardly surprising as it hardly seems to recognise some OpenCL streams at all :-).

Essentially it is meant to convert an OpenCL structure into an image and it can do a pretty good job of dealing with video and streamed movies too.

This is a bit of a tangent to my creating some really effective OpenCL Optical Flow kernel based constructs. That is definitely a work in progress.

This little beauty shows how the Colours structure can be turned into a representation of the video [Coil.mov - online version]. There actually is more variation amongst the other structure members than this grab gives lie to :-) .

The other OpenCL structures also yield graphically renderable information, chromatic and movement. Note the image output from the Convertor patch. The other two Convertors output the rainbow background and the texture grid.

Put together, they make for a rather graceful visualizer, albeit a slow vehicle fps wise, which I don't find matters too much, especially as feeding the audio peak to the movie rate [Asynchronous mode] keeps the whole moving structure synced to the audio throughput.

Still it would be nice to find an even more effective fps capable kernel based construct.

Hey everyone! I would really appreciate any help!!!

I'm new to Quartz Composer and need to make a composition that I believe is quite simple- but I don't really know what I'm doing. I am trying to take realtime video from an external USB webcam (or my iSight, if that is easier) and put it on a delay of as long as possible. I think I read somewhere that 20 seconds will be about the maximum because it takes a lot of computing power or something. So really, all i need is to display the video input on a slight delay. if someone could take a screencapture or something of how to set this up that would be amazing.

I am running on a dual core MacBook Pro OS X 10.5.8

Oh.... and I have to figure this out by tomorrow morning.... :/ THANKS!

(bumped to a new thread instead of just a comment on a previous topic...)

I'm trying to get a Point Grey FireFly MV FFMV-03M2C firewire color camera working... Kineme Video Device Info is reporting the device is there and capable, using the CameraExplorer qtz shows the available settings, but no image ever is output. Looking at the console, I'm just seeing the camera failing to handshake due to some sort of color coding issue.

Apr 12 12:48:41 *: [VTVideoInputPatch execute:time:arguments:]  Waiting for transmission to begin...
Apr 12 12:48:40 *: libdc1394 error: Format_7 Error_flag_1 is set: in _dc1394_v130_handshake (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 122): invalid image position, size, color coding or ISO speed
Apr 12 12:48:40 *: libdc1394 error: Format_7 Error_flag_1 is set: in dc1394_format7_set_roi (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 803): Handshaking failed after setting color_coding
Apr 12 12:48:41 *: libdc1394 error: Format_7 Error_flag_1 is set: in _dc1394_v130_handshake (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 122): invalid image position, size, color coding or ISO speed
Apr 12 12:48:41 *: libdc1394 error: Format_7 Error_flag_1 is set: in dc1394_format7_set_roi (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 803): Handshaking failed after setting color_coding

or similar errors:

Apr 12 12:57:41 *: libdc1394 error: Format_7 Error_flag_1 is set: in _dc1394_v130_handshake (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 122): invalid image position, size, color coding or ISO speed
Apr 12 12:57:41 *: libdc1394 error: Format_7 Error_flag_1 is set: in dc1394_format7_set_roi (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 803): Handshaking failed after setting color_coding
Apr 12 12:57:48 *: libdc1394 error: Format_7 Error_flag_1 is set: in _dc1394_v130_handshake (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 122): invalid image position, size, color coding or ISO speed
Apr 12 12:57:48 *: libdc1394 error: Format_7 Error_flag_1 is set: in dc1394_format7_set_roi (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 803): Handshaking failed after setting color_coding
Apr 12 12:57:52 *: libdc1394 error: Format_7 Error_flag_1 is set: in _dc1394_v130_handshake (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 122): invalid image position, size, color coding or ISO speed
Apr 12 12:57:52 *: libdc1394 error: Format_7 Error_flag_1 is set: in dc1394_format7_set_roi (/Users/smokris/prj/qc/plugin/VideoTools/libdc1394/dc1394/format7.c, line 803): Handshaking failed after setting color_coding

This is after I've tried all possible resolutions and framerates, and thoroughly explored all color coding formats both manufacturer-documented and Kineme-suggested (MONO8, MONO16, RAW8, RAW16, YUV411, YUV422, YUV444, RGB8, Y8, Y16) and when I'm not getting "Invalid Color Coding" responses, I'm seeing "Waiting for transmission to begin..." and then nothing...

I believe this camera only supports monochrome images in its standard non-Format 7 modes (with a QuickTime based video feed, I only see a monochrome image). Unfortunately, I need color output.

The documentation for the camera suggests that color data is not retained when binning is used ("Color binning is not supported on the FMVU-03MTC or FFMV-03M2C. Color data is not maintained when operating these models in pixel binning mode."). So I guess that means if we use a Format 7 mode, we have to define the ROI but I'm not sure how? Kind of beyond my expertise by this stage.

So no support yet using Kineme Video Input - any further suggestions?

Hi everybody,

I've been working on some (really basic :) augmented reality stuff for the last few days which I want to be presented in a POS style fashion, e.g. a portrait oriented display with a mounted, external (USB) camera.

For some reason the Quartz Composer "Video Input" patch captures the signal from external USB video class (UVC) cameras in their native video size, for instance 1600x1200 (high-end OSX Logitech something) or 1280x960 (XBox Camera) pixels. Which - as this is obviously an uncompressed YCrCb 4:2:2 video stream - easily saturates the USB bus, which leads to slow frame rates and/or image tearing. As UVC is generally supported in OSX since 10.4.9 there is no prefs app or setting to reduce the capture size. The image of the built-in iSight is captured at 640x480 pixels in QC, even though the iSight has a higher native video size (1280x1024 if I remember correctly).

So, pretty long story, short question.

Is there an alternative video input patch where you can modify the capture/output size?

The QC built-in "Video Input" patch once had such settings but unfortunately this was removed in QC3/OSX 10.5, according to the QC3 release notes.

Thanks in advance. Take care, Oli

Hi there, I'm a bit new to QC but I already developped some apps & effects and really enjoy it. One problem I have: I did an application for light painting using video input (http://mattbeghin.net/?p=55). The problem I have is that it doesn't seem to be possible to select the video input without modifying the quartz composition, which is not a solution for me since I'm not on site each time it is used, and the camera is not always the same. Is there any Video Input patch that has an input parameter to change the device used ? Thanks ! Matt