Blog for work on my Masters thesis - a survey of methods for evaluating media understanding, object detection, and pattern matching algorithms. Mostly, it is related to ViPER, the Video Performance Evaluation Resource. If you find a good reference, or would like to comment, e-mail viper at cfar.umd.edu.
Archives
Media Processing Evaluation Weblog
Monday, January 26, 2004
Vista: A Software Environment for Computer Vision Research
Vista presents a set of unix CLI utilities for manipulating images and other related data structures. The software seems to have stopped developed not long after the publication of the paper, although it seems to have a few followers still. Its integrated metadata format provides a few features missing from ImageMagick shell scripts.
Link Reference
@inproceedings{ pope94vista, author = "Pope, A.R. and Lowe, D.G.", title = "Vista: A Software Environment for Computer Vision Research", booktitle = "CVPR94", pages = "768-772", year = "1994", url = "citeseer.nj.nec.com/pope94vista.html" }
- posted by David @ 7:47 PM
Wednesday, January 21, 2004
Building an OS X Application in Eclipse
I found a helpful ant task for building mac.apps, called the Apache Ant Jar Bundler. It is pretty useful, and available under the GPL.
- posted by David @ 4:42 PM
Sunday, January 18, 2004
The G5 Powermac
So, I've started using the dual g5 machines in the WAM labs recently. These 2x2gHz machines smoke the single 1.7 p3 I have on my desk. I can actually decode video at a decent framerate with jmpeg! The time test, which usually doesn't finish in real time despite not going through the hassle of actually displaying the video, did finish in real time for the small sample video, if going about a little less than three quarters real time. The results:
.Averaged 35.527082119976704fpms .Averaged 21.779144715927895fpms F Time: 776.975 There was 1 failure: 1) testDuckAndCover (edu.umd.cfar.lamp.mpeg1.test.VideoPerformanceTest) junit.framework.AssertionFailedError: Video didn't decode close enough to real time. Took 764401 ms, but should be less than 555488 ms.
As there are people out there who use the platform, I probably should start making the application more mac-friendly, including moving the menu to the top, and setting up the icon, per sun's advice. There are also a few problems with the table widget drawing that indicate some problems with how I'm writing my cell renderers/editors. I've also got the dark-red dot showing up in the close button and I've made the scroll bars flush with the window edges when appropriate, something that Safari still has trouble with. There is still plenty to do, though.
- posted by David @ 4:19 PM
Wednesday, January 14, 2004
MPEG-2 Support
With Tony working on MPEG-2 support, I'm focussing my attention on other matters, namely fixing the schema editor and adding editing to the timeline. It appears that there won't be any real evaluation of VACE Phase II until next year, so the focus this year will be on the annotation tools and defining the metrics. Hopefully, ViPER will be at the center of a lot of this.
- posted by David @ 4:42 PM
Wednesday, January 07, 2004
SourceForge Ant Task for the File Release System
I've recently replaced the makefiles and shell scripts for generating the viper distribution files with a set of ant build.xml files. In order to make the process even easier, there is an ant task for using the SourceForge FRS, which I'll be figuring out how to use shortly.