Since I've been doing so much online surfing for solutions to my server and CPU issues, I thought I should post some of the links that have been particularly helpful with some notes on the information.
+ Getting beyond 50% CPU usage in Jitter
This C74 thread gives some helpful solutions on getting Jitter to utilize multiple processor cores. I'm interested in the concept of running one patch in Max and the other in Max Runtime, so that the computer recognizes them as two separate processes and will distribute them how it will.
+ FPS Issues with Upgrade
This thread discusses the sudden drop in FPS when upgrading to Max 4.6 and Jitter 1.6. There is apparently an message that can be sent to Jitter to tell it how many cores and threads to use, as well as an option that defaults to "on" when it didn't used to.
+ Jitter Message Reference
A page listing all the messages that can be sent to the object Jitter to alter the behavior of Jitter as a whole. Good to know!
+ CheckUpDown Error Guide
CheckUpDown.com has a page for every HTTP error code that details common causes for the error as well as possible fixes. Useful for figuring out exactly what problem the web servers are having with Silo.
+ Enabling PHP on your Apache Server
Self-explanatory. Was very easy to follow.
+ Apache Webserving with Mac OS X
A 5-part guide with tons of information on using the Apache server that comes with OS X.
- M
Thursday, May 31, 2007
Non-Progress
Well, it looks like things are at a standstill for the moment. Seems as though both the school's server and my advisor's web server have different issues with Silo that we can't solve ourselves. Zero Linden, the creator of Silo, has been a BIG help in getting us through most of our problems, but now it's all in the hands of the school's IT department. It looks like the problem with the school's server is that Silo requires access to the DELETE method to function properly, and that ability is likely disabled for very good reasons. Hopefully they'll be willing to let us poke through some of their heavy security so Silo can do what it needs to do.
My video tracking patch is still having bizzare CPU usage issues. Though the metro is driving most parts of the patch at 100ms, I'm hearing notes anywhere from once every 500ms to once per second. That is, of course, variable depending on how hard the CPU is working. Often times, I get an odd sort of galloping effect. The strange part is that this seems to happen no matter how much CPU is being used by the online community interface. This being a dual-core processor, I usually end up with Max/MSP using anywhere from 65-95%, depending on which interface I use. SL usually takes up no less than 95% (causing Max to use around 70%), while World of Warcraft uses only ~35% (causing Max to use ~85/90%). Notes seemed to come less often, if anything, running Warcraft, presumably because it is more populated and there are more people to track. One would think that with more CPU available to it, Max would take advantage and be able to process everything being asked of it. Rather, it seems to be restricting itself to just one core and scaling based on that. Perhaps the cv.jit objects don't support multithreading? I'm going to go ahead and build a polyphonic synthesizer to insert into the patch and see how the CPU behaves after that point. I suppose if that doesn't go well then I'll have to look into using two computers for this portion of the project.
- M
My video tracking patch is still having bizzare CPU usage issues. Though the metro is driving most parts of the patch at 100ms, I'm hearing notes anywhere from once every 500ms to once per second. That is, of course, variable depending on how hard the CPU is working. Often times, I get an odd sort of galloping effect. The strange part is that this seems to happen no matter how much CPU is being used by the online community interface. This being a dual-core processor, I usually end up with Max/MSP using anywhere from 65-95%, depending on which interface I use. SL usually takes up no less than 95% (causing Max to use around 70%), while World of Warcraft uses only ~35% (causing Max to use ~85/90%). Notes seemed to come less often, if anything, running Warcraft, presumably because it is more populated and there are more people to track. One would think that with more CPU available to it, Max would take advantage and be able to process everything being asked of it. Rather, it seems to be restricting itself to just one core and scaling based on that. Perhaps the cv.jit objects don't support multithreading? I'm going to go ahead and build a polyphonic synthesizer to insert into the patch and see how the CPU behaves after that point. I suppose if that doesn't go well then I'll have to look into using two computers for this portion of the project.
- M
Thursday, May 24, 2007
Progress: Fixes and Issues
The good news is that Silo is now working! The bad news is that I had it right all along and the university was just blocking my apache server. Three hours of frustration for a simple fix, but at least it works now. I also read through both the code for the chat logger and the code needed to send data out of SL and into Silo, and it looks like the changes needed are pretty straightforward. I hope to have the chat logger working and sending data to Silo by the end of today. My advisor has been investigating the other end of the data chain, getting data from PHP to Max/MSP via OpenSound Control, and that has been going well. Data is flowing through the chain, though we are still figuring out how to format the messages so we get the result we want from Max. All in all, the chat logging trial has been going well, and I hope to have a working proof of concept by the end of next week.
The video tracking has been moving along, but I'm afraid I may have hit an impasse. Despite the fact that I'm doing this work on a 2.33GHz Intel Core 2 Duo processor, I'm almost maxed out on CPU power while running both Second Life and my video tracking patch. I know the major problems are jit.desktop and cv.jit.faces, but I'm not sure what to do with them to reduce the CPU strain. The video is split into four quadrants, and right now the face tracking and demo audio are only running on one of the four quadrants. If the CPU is already near maxed out, then it would probably kill my computer to have all four quadrants tracking faces with full audio effects resulting. I've already disabled the Use Onscreen option for the jit.pwindow objects, inserted some speedlim objects, and taken out a few redundant things; I'm not sure what else I can do. I'm going to go through the Cycling '74 forums to look for advice, but I'm really afraid that it's going to come down to needing two computers: one for tracking and one for audio. I'm very opposed to that option, since one of my main goals was to make this project as mobile as possible, but if that's what it takes then so be it. We'll see if it needs to come to that.
- M
The video tracking has been moving along, but I'm afraid I may have hit an impasse. Despite the fact that I'm doing this work on a 2.33GHz Intel Core 2 Duo processor, I'm almost maxed out on CPU power while running both Second Life and my video tracking patch. I know the major problems are jit.desktop and cv.jit.faces, but I'm not sure what to do with them to reduce the CPU strain. The video is split into four quadrants, and right now the face tracking and demo audio are only running on one of the four quadrants. If the CPU is already near maxed out, then it would probably kill my computer to have all four quadrants tracking faces with full audio effects resulting. I've already disabled the Use Onscreen option for the jit.pwindow objects, inserted some speedlim objects, and taken out a few redundant things; I'm not sure what else I can do. I'm going to go through the Cycling '74 forums to look for advice, but I'm really afraid that it's going to come down to needing two computers: one for tracking and one for audio. I'm very opposed to that option, since one of my main goals was to make this project as mobile as possible, but if that's what it takes then so be it. We'll see if it needs to come to that.
- M
Wednesday, May 23, 2007
Progress: Face Recognition
My motion tracking patch now recognizes faces. I've programmed it so that it feeds the coordinates of the face into cv.jit.track and plays notes based on the movement of the face. I also threw in a quick fix so that there wouldn't be silence when the face is lost, but I'll be taking it out and replacing it with something more elegant. Everything has been tested within SL and so far, so good.
Though I've had mostly success with this patch so far, I believe I'll be grinding to a halt before too much longer. I've been having CPU usage issues lately, and the patch doesn't even contain all the elements it needs to yet. I know my two main issues are jit.desktop and cv.jit.faces, which the help file states contains a very CPU-intensive algorithm. The patch certainly needs a lot of cleanup, both for CPU reasons and for aesthetic reasons, so I that will be one of my focuses for today. I'd also like to see if I can figure out how to incorporate the LSL script that allows communication with Silo into the chat logger, despite the fact that Silo still isn't up on the apache server. Between those two, I think I'll have plenty to do today.
- M
Though I've had mostly success with this patch so far, I believe I'll be grinding to a halt before too much longer. I've been having CPU usage issues lately, and the patch doesn't even contain all the elements it needs to yet. I know my two main issues are jit.desktop and cv.jit.faces, which the help file states contains a very CPU-intensive algorithm. The patch certainly needs a lot of cleanup, both for CPU reasons and for aesthetic reasons, so I that will be one of my focuses for today. I'd also like to see if I can figure out how to incorporate the LSL script that allows communication with Silo into the chat logger, despite the fact that Silo still isn't up on the apache server. Between those two, I think I'll have plenty to do today.
- M
Tuesday, May 22, 2007
Progress: First Noise
The motion tracker makes sound! Right now it's little more than a crude noisemaker, but I do indeed have sound coming from motion. Obviously a step in the right direction; however, the motion tracking still needs to be refined. Today will mostly be focused on incorporating either cv.jit.faces or cv.jit.features2track. I'm hoping that one of these two objects will provide more reliable tracking within SL than cv.jit.track does. If all goes well for the next few hours (or if I get frustrated and need a break), I may work on the MSP side of the patch. The short-term goal is a simple polyphonic synthesizer, and I'll add more features as the project progresses.
Still no luck getting Silo installed on my apache server. Time to go forum diving again, it seems.
- M
Still no luck getting Silo installed on my apache server. Time to go forum diving again, it seems.
- M
Monday, May 21, 2007
Progress: Fun with PHP
Thursday the 17th proved to be a very frustrating day indeed. I decided to focus on the issue of getting data from Second Life into MaxMSP by way of my laptop's apache server. Enabling the server is baby simple; a checkbox in system preferences. Enabling PHP on the server was somewhat more complicated, but this Very Helpful Tutorial on editing your apache configuration files in Terminal worked out perfectly. Tested, it works, all is well.
Now, on to the real issue: Getting Silo up and running.
Silo was created by Zero Linden for storing and retrieving data using HTTP. Silo comes in the form of a PHP file that needs to be installed onto the apache server. This is where my problems have erupted. Where to install the file depends on your apache configuration. It seems to me (with my very limited knowledge) that the file should be placed under the Document Root, which for Mac OS X is /Library/WebServer/Documents. I tried testing it using the small python program that accompanies Silo, and the test came up with an error.
The other two most common apache configurations require that silo be placed either in the user's public-html directory or outside the document tree. From reading through my apache configuration, I don't believe either of these are it, though I admittedly have little to no experience in these matters. Of course, the Silo ReadMe does state that there are other, more complicated apache configurations that he did not list. If that ends up being the case, I'll need to track someone down who is knowledgeable about these things.
Needless to say, after three hours of trying to puzzle this out, I was frustrated beyond belief. Time to move on to something else!
I began constructing my motion tracking patch using the cv.jit objects. I ran a few tests, seeing how they would track movement in SL, and everything seems to be fine so far. Currently, the issue is incorporating cv.jit.faces into the mix to track individual people. It doesn't seems to output any sort of coordinates for the location of the face, but I may be able to use the output of the draw object used to outline the faces somehow. I'm still looking into it, and hopefully I'll have an answer by the end of today or tomorrow.
Onward!
- M
Now, on to the real issue: Getting Silo up and running.
Silo was created by Zero Linden for storing and retrieving data using HTTP. Silo comes in the form of a PHP file that needs to be installed onto the apache server. This is where my problems have erupted. Where to install the file depends on your apache configuration. It seems to me (with my very limited knowledge) that the file should be placed under the Document Root, which for Mac OS X is /Library/WebServer/Documents. I tried testing it using the small python program that accompanies Silo, and the test came up with an error.
The other two most common apache configurations require that silo be placed either in the user's public-html directory or outside the document tree. From reading through my apache configuration, I don't believe either of these are it, though I admittedly have little to no experience in these matters. Of course, the Silo ReadMe does state that there are other, more complicated apache configurations that he did not list. If that ends up being the case, I'll need to track someone down who is knowledgeable about these things.
Needless to say, after three hours of trying to puzzle this out, I was frustrated beyond belief. Time to move on to something else!
I began constructing my motion tracking patch using the cv.jit objects. I ran a few tests, seeing how they would track movement in SL, and everything seems to be fine so far. Currently, the issue is incorporating cv.jit.faces into the mix to track individual people. It doesn't seems to output any sort of coordinates for the location of the face, but I may be able to use the output of the draw object used to outline the faces somehow. I'm still looking into it, and hopefully I'll have an answer by the end of today or tomorrow.
Onward!
- M
Wednesday, May 16, 2007
Progress: Video Tracking
Step One: What externals to use?
After several failed attempts at downloading and installing SoftVNS (corrupted files?), I finally gave up and began exploring other motion tracking options. My forum diving over at Cycling '74 turned up three more possible solutions: Electrotap's TapTools, Eric Singer's Cyclops, and Jean-Marc Pelletier's cv.jit library. Surface analysis of the three suggests that the cv.jit library will likely be my best bet, based on extra features and budget considerations (it's free).
After spending some time with the cv.jit library's well-designed help files, I've found both the objects I need for motion tracking and some extra goodies that should prove useful. For instance, cv.jit.faces tracks the presence of a human face in a video. I paired it up with jit.desktop to see if it would recognize animated faces in Second Life and, lo and behold, it does. This opens up all sorts of possibilities as far as giving individual users control over specific parameters or their own personal track. While there is no cv.jit object specifically for color tracking, the suckah object should serve well enough.
More fiddling is forthcoming.
After several failed attempts at downloading and installing SoftVNS (corrupted files?), I finally gave up and began exploring other motion tracking options. My forum diving over at Cycling '74 turned up three more possible solutions: Electrotap's TapTools, Eric Singer's Cyclops, and Jean-Marc Pelletier's cv.jit library. Surface analysis of the three suggests that the cv.jit library will likely be my best bet, based on extra features and budget considerations (it's free).
After spending some time with the cv.jit library's well-designed help files, I've found both the objects I need for motion tracking and some extra goodies that should prove useful. For instance, cv.jit.faces tracks the presence of a human face in a video. I paired it up with jit.desktop to see if it would recognize animated faces in Second Life and, lo and behold, it does. This opens up all sorts of possibilities as far as giving individual users control over specific parameters or their own personal track. While there is no cv.jit object specifically for color tracking, the suckah object should serve well enough.
More fiddling is forthcoming.
Grant Proposal Abstract
I suppose the best introduction to my research is to read the abstract for the proposal I wrote back in January. Anyone interested in reading the entire proposal is welcome to e-mail me.
Abstract
In recent years, interactive music has emerged as a way of involving those without musical training in music composition and performance. The goal of this project is to investigate a new medium for interactive music, the online graphical community. Using the programming environment Max/MSP/Jitter, I will build three small computer programs to convert text and video data from the users of the online community into music. This allows people from all over the world to create music together, with or without any prior musical training. The final product will be an interactive real-time performance blending the physical performance space with the virtual one, both visually and aurally.
Abstract
In recent years, interactive music has emerged as a way of involving those without musical training in music composition and performance. The goal of this project is to investigate a new medium for interactive music, the online graphical community. Using the programming environment Max/MSP/Jitter, I will build three small computer programs to convert text and video data from the users of the online community into music. This allows people from all over the world to create music together, with or without any prior musical training. The final product will be an interactive real-time performance blending the physical performance space with the virtual one, both visually and aurally.
Subscribe to:
Posts (Atom)
