Jitter + IP camera and or syphon?
Ok after a lot of dead end's in former topics I'm hoping to get a working idea of how to approach streaming IP camera's.
So first I've looked at the ip camera from VLC, so syphon is a way, but the syphon kit for vlc to max is gone from the internet... (dramatic sound)
Then I looked inside maxmsp/jitter with jit.uldl but this is not constant in the way it downloads and shows the feed and has much latency and skips even more frames. Whilst VLC doesn't.
Example.
Then I've read lots of jit.qt.movie with read rtsp://(link of the camera) = rtsp://110.4.60.239:80/-wvhttp-01-/GetOneShot?image_size=640x480&frame_count=no_limit
But this doesn't give me any image, nor do I find anything concrete on how to use the rstp comment.
Then I read something very old of jit.broadcast and jit.qt.broadcast but this object doesn't seem to be working anymore or at least nog on my version of max 7
I generally don't care about the wole latency of the stream as long as it as fluent as possible when showed on the jit.qt.movie ( equal to the max possible quality e.g. like VLC). I also do not think its my network that's slowing it down.
i would go the syphon route. i would use something like this:
https://github.com/bakercp/IPCAM2SYPHON
syphon objects for Max can be installed via the Package Manager.
I have a lot of experience working with IP cameras in Jitter. Rob is right, the SyphonNetCamera app work well and is very stable, but provides scarce visual feedback that it is working. IPCAM2SYPHON provides a preview window and fps/Kbps statistics, but is not stable and will crash intermittently.
I usually use IPCAME2SYPHON to set up the shot and make sure everything's working, and then run SyphonNetCamera for the actual work.
The most challenging thing working with IP cameras is that most manufacturers don't publish the URL for their MJPEG stream, which both apps above will need as they can't receive H.264. I've posted about cameras I know work as I've used them quite a bit - see below.
Hi, I found a way to show a stream thru syphon and vlc
I have this patch now. The syphon spits out the video in the jit.window clientcontext. I trying to get a motion detection from that window.
But this doesn't work. I've linked it with a jit.qt.movie to get image movement recognition from it but it doesn't work.
Ok putting a matrix between it and connecting it to the luma seemed to do the trick.
First small question, why does the jit.matrix flip(s) my image and how to bring this back to normal.
Then second despite my pretty powerful Imac, the patch has running problems and as in the future it will need to run a couple of these patches simultaneously I'm wondering what to clean out to make it as light processable as possible. Can I for example remove this part
to have it run smoother (If I do it know It doesn't matter because it only runs one ip). But the jit.gl.videoplane, jit.gl.render are those 'heavy' objects?
How do I install SyphonNetCamera? I don't get how the github files work?
Though I have knowledge of how githubs work for other types of platforms. I don't understand this one.
the heavy parts are reading back to matrix from texture, and the cv processing. you should make you matrices as small as possible that can still do the tracking, e.g. 80 x 60 may be all that's needed.
to flip the image, you can send it to a jit.gl.slab object prior to the jit.matrix conversion, and use the @rect attribute to flip. you can also resize in the jit.gl.slab: [ jit.gl.slab @rect 0 1 1 0 @adapt 0 @dim 80 60 ]
this flips the image vertically and resizes it to 80 x 60
awesome no more lag for the moment.
On on other note, is it normal when you have two jit.gl.slab that one rotates back upsidedown? The first one was normal but when I attached another jit.gl.slab to a new matrix, the first one flipped back upside down?
Hello all,
thanks for all of the above ^^
I've been working on a project where I am taking Department of Transportation IP feeds and running them through some custom CV patches to motion detect and object recognize. I am struggling right now over the best way to get the IP cams into my patch. I've been using Vade's Security Player tool : http://abstrakt.vade.info/?p=80 with a few tweaks and it sort of works, but I am getting a ton of errors and I wont even begin to pretend to know what I am doing when it comes to java. I am essentially trying to create a large grid out of these feeds, and when I get too many of the bpatchers in there, I get java heap space errors. I am wondering if there is a more robust way to do this? Would a syphon be better? I looked at the IPCAM2SYPHON tool, but the problem is the DOT feeds are in .jpg form, and need to constantly reconnect to update. The IPCAM2SYPHON stops after 20 reconnects. VLC isnt ideal because I am working with such a large number of feeds. Here is a snippet of how I am inputting the IP cams and gluing them together currently. Any suggestions?
Thinking of a 360 vr video output setup. I guess need those panoramic cams.