Join us now at the IRC channel.
| schestowitz | [23:54] <tessier> What do these rss extensions do? | Jul 09 01:17 |
|---|---|---|
| schestowitz | [23:55] <schestowitz> They fetch data from the RSS feed of the site and then output that as wiki content | Jul 09 01:17 |
| schestowitz | [23:55] <schestowitz> The same problem existed before, twice even, and you did something on the server which made it work. I don't know what exactly. | Jul 09 01:17 |
| schestowitz | [00:01] <tessier> Oh...they connect out to some other site? | Jul 09 01:17 |
| schestowitz | [00:02] <schestowitz> Not AFAIK, but I didn't look at the code | Jul 09 01:17 |
| schestowitz | [00:03] <tessier> They retrieve the rss url from the site and then output it as wiki content? | Jul 09 01:17 |
| schestowitz | [00:03] <schestowitz> One time this extension broke was when varnish was installed, IIRC | Jul 09 01:17 |
| schestowitz | [00:03] <schestowitz> Yes, and there is also caching | Jul 09 01:17 |
| schestowitz | [00:03] <schestowitz> to avoid loading up the feed each time the front page is refreshed | Jul 09 01:17 |
| schestowitz | [00:07] <tessier> So it is trying to retrieve the rss url from the same machine on which that script is running... do you tell it exactly what the url for the rss feed is? What url is it? | Jul 09 01:17 |
| schestowitz | [00:08] <schestowitz> I tried some variations of it to overcome the problem Right now it's accessing http://techrights.org/feed/atom/ - the original URL it was always accessing. | Jul 09 01:17 |
| schestowitz | [00:09] <schestowitz> I can wget that | Jul 09 01:17 |
| schestowitz | [00:10] <tessier> What is the error that you are seeing? You are trying to wget it from the server itself, correct? | Jul 09 01:17 |
| schestowitz | [00:11] <schestowitz> Yes, from the server. In the wiki I can't quite see the error, I didn't look at PHP error logs | Jul 09 01:17 |
| schestowitz | [00:13] <schestowitz> tail -n600 /var/log/httpd/techrights.org-error_log doesn't show anything related to this | Jul 09 01:17 |
| TechrightsBot | Not a web page! Aborting application/atom+xml; charset=UTF-8 type | Jul 09 01:18 |
| schestowitz | > Dietrich is a good example of what an open source advocate should not be: he | Jul 09 02:58 |
| schestowitz | > has low social skills, he wants things for free (or to have people pay him | Jul 09 02:58 |
| schestowitz | > to keep his hobby going - and uses extreme language in his pleas), he | Jul 09 02:58 |
| schestowitz | > belittles others who disagree with him, he goes against his own stated | Jul 09 02:58 |
| schestowitz | > morals and deletes non-offensive comments from his site, etc. | Jul 09 02:58 |
| schestowitz | > | Jul 09 02:58 |
| schestowitz | > There are many very good people in the open source community. I disagree | Jul 09 02:58 |
| schestowitz | > with much of what you say but still think your morals are mostly good (you | Jul 09 02:58 |
| schestowitz | > do not beg for money, do not generally attack people, etc.). Sadly you are | Jul 09 02:58 |
| schestowitz | > one of the few COLA "advocates" who is not essentially morally bankrupt... | Jul 09 02:58 |
| schestowitz | > and it seems Dietrich is of the same cloth as most of the COLA "advocates". | Jul 09 02:58 |
| schestowitz | > Hello Dr Schestowitz, | Jul 09 04:46 |
| schestowitz | > | Jul 09 04:46 |
| schestowitz | > Firstly, a big thank you for your videos detailing the process of | Jul 09 04:46 |
| schestowitz | > opencv_createsamples, opencv_traincascade and detectMultiScale. They | Jul 09 04:46 |
| schestowitz | > have come in very handy with helping me understand the workflow pipeline. | Jul 09 04:46 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > But I have hit a bit of a snag and would be very grateful for some | Jul 09 04:47 |
| schestowitz | > guidance. I eventually want to be able to detect European Starlings in | Jul 09 04:47 |
| schestowitz | > my fig tree using OpenCV on a Raspberry Pi, so that I can trigger a | Jul 09 04:47 |
| schestowitz | > scaring device to keep the buggers from eating my figs! As a baby step, | Jul 09 04:47 |
| schestowitz | > I'm starting small with a star-shaped cookie cutter with a mere 11 | Jul 09 04:47 |
| schestowitz | > positive samples and 20 negative samples (one of each is attached). | Jul 09 04:47 |
| schestowitz | > I've successfully created the 20th-stage cascade.xml (also attached), .... | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > ...but when I run detectMultiScale on a 320x240 video frame it takes | Jul 09 04:47 |
| schestowitz | > nearly 1 minute to process a single frame! The code for the call is: | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > cascade.detectMultiScale(gray_image, matches, 1.1, 5, | Jul 09 04:47 |
| schestowitz | > CV_HAAR_SCALE_IMAGE, Size(80,80)); | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > I've searched the web and can't find any strong guidance on this aside | Jul 09 04:47 |
| schestowitz | > from this one post: | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > http://answers.opencv.org/question/4722/cascadeclassifierdetectmultiscale-takes-a-minute/ | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| TechrightsBot | Title: CascadeClassifier.detectMultiScale takes a minute! Why? - OpenCV Q&A Forum .::. Size~: 37.41 KB | Jul 09 04:47 |
| schestowitz | > ...which makes vague mentions of using a better classifier, more | Jul 09 04:47 |
| schestowitz | > positive/negative samples, increased number of neighbors. | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > What would you recommend? I'd love to get some realtime (at least 1fps) | Jul 09 04:47 |
| schestowitz | > detection working on a simple classifier without investing several hours | Jul 09 04:47 |
| schestowitz | > in preparing a temporary test classifier for the cookie cutter. | Jul 09 04:47 |
| schestowitz | > | Jul 09 04:47 |
| schestowitz | > Any / all technical suggestions are hugely appreciated! | Jul 09 04:47 |
| schestowitz | Hi Ken, | Jul 09 04:47 |
| schestowitz | It has been nearly a year since I last dealt with this task, but from what I can recall, on a single core ARM processor on a tablet I would require 6 seconds or so per frame if I did things poorly, maybe 6 FPS if done correctly. I think that a multi-resolution approach is already part of the algorithm, so scaling things would not change much. I don't think that size of the training set matters, except when you create and refine a | Jul 09 04:47 |
| schestowitz | cascade classifier. | Jul 09 04:47 |
| schestowitz | Try working with some demo like the face recogniser (OpenCV demo classic, at least for Android), using the cascade they provide, then see how it performs on a RaspberryPi. That would give some baseline... | Jul 09 04:47 |
| schestowitz | HTH, | Jul 09 04:47 |
| schestowitz | "There is no emergency where broken communications are helpful. Denial of communications only helps bullies get away with hurting people." | Jul 09 04:47 |
| schestowitz | > Wow, thanks for the quick response! I actually did proceed just as you | Jul 09 04:55 |
| schestowitz | > suggest - first with a face detection sample which runs about 5fps or | Jul 09 04:55 |
| schestowitz | > so, then replaced the cascade XML file which introduced the performance | Jul 09 04:55 |
| schestowitz | > issue. There are still some things I can do to try to pin down the | Jul 09 04:55 |
| schestowitz | > issue, but I have a feeling lots of experimentation lies ahead :) | Jul 09 04:55 |
| schestowitz | > | Jul 09 04:55 |
| schestowitz | > Ken | Jul 09 04:55 |
| schestowitz | Check out the size, in bytes, of the two cacades. | Jul 09 04:55 |
| schestowitz | > Hm, the lbpcascade_frontalface.xml is 52KB, and mine is 361KB. I'm not | Jul 09 05:12 |
| schestowitz | > sure what to make of the discrepancy though -- do you think I need to | Jul 09 05:12 |
| schestowitz | > tune mine to be smaller, and if so, what setting might I need to adjust? | Jul 09 05:12 |
| schestowitz | I can't recall the settings, but the classifier can be made smaller. | Jul 09 05:12 |
| schestowitz | troll: | Jul 09 05:21 |
| schestowitz | https://twitter.com/ManningTrial/status/354390794678964225 | Jul 09 05:21 |
| TechrightsBot | Not a tweet? | Jul 09 05:21 |
| schestowitz | "educate me Herr Doktor. Explain how #Manning trial has been "like lynching." I didn't realize KKK involved—but then you'd know.' | Jul 09 05:21 |
| schestowitz | what a douchebag https://twitter.com/ManningTrial/status/354390794678964225 | Jul 09 05:23 |
| TechrightsBot | Not a tweet? | Jul 09 05:23 |
| schestowitz | I think he is ANTI Manning https://twitter.com/ManningTrial/status/354408320536297472 | Jul 09 05:23 |
| TechrightsBot | Not a tweet? | Jul 09 05:23 |
| schestowitz | "actually, passing state secrets to the enemy has ALWAYS been considered treasonous; so #Manning conviction won't change that." | Jul 09 05:23 |
| schestowitz | https://twitter.com/ManningTrial/status/354384210431197184 | Jul 09 05:24 |
| TechrightsBot | Not a tweet? | Jul 09 05:24 |
| schestowitz | "Ellsberg & #Manning = apple & orange in one supremely important way—Ellsberg was a civilian; Manning a soldier. Invalidates comparison." | Jul 09 05:24 |
| schestowitz | BS | Jul 09 05:24 |
| schestowitz | Ellsberg worked for the gov | Jul 09 05:24 |
| schestowitz | https://twitter.com/ManningTrial/status/354323139854995458 | Jul 09 05:25 |
| TechrightsBot | Not a tweet? | Jul 09 05:25 |
| schestowitz | "Coombs showing Apace video in court = inflammatory stunt. Whoever thinks Judge Lind never before saw "Collateral Murder" is Fool 4 #Manning." | Jul 09 05:25 |
| schestowitz | https://twitter.com/ManningTrial/media/grid | Jul 09 05:31 |
| TechrightsBot | Title: ManningTrial (ManningTrial) on Twitter .::. Size~: 191.76 KB | Jul 09 05:31 |
| schestowitz | LOL: https://d2tq98mqfjyz2l.cloudfront.net/image_cache/1373342587680175_animate.gif | Jul 09 05:35 |
| TechrightsBot | Not a web page! Aborting image/gif type | Jul 09 05:35 |
| schestowitz | > Ok, I'm trying again with 24x24 positives (vs former 50x50). I've also | Jul 09 05:39 |
| schestowitz | > found some in-depth docs about other training parameters, will plug away | Jul 09 05:39 |
| schestowitz | > at it, thanks for the tip. | Jul 09 05:39 |
| schestowitz | http://schestowitz.com/royrianne/gallery/index.php/ | Jul 09 08:14 |
| schestowitz | Gallery now has four new improvements: menu at the top, number of photos in each album is shown as well as other additional description, in image view there is a 3x3 matrix of related album contents, and finally some translucent navigation buttons on top of images. | Jul 09 08:14 |
| TechrightsBot | Title: Roy and Rianne's Wedding and Beyond .::. Size~: 61.18 KB | Jul 09 08:14 |
| schestowitz | > Drat, that didn't make any difference. I tried doing 30-stage training, | Jul 09 08:28 |
| schestowitz | > which made the file significantly bigger (541 KB) then 20-stage which | Jul 09 08:28 |
| schestowitz | > was almost the exact same size as before (360 KB). Will do more | Jul 09 08:28 |
| schestowitz | > research and trials. It's a shame this isn't all documented better :^/ | Jul 09 08:28 |
| *MinceR has quit (*.net *.split) | Jul 09 10:09 | |
| *libertybox_ has quit (*.net *.split) | Jul 09 10:11 | |
| *pidgin_log has quit (*.net *.split) | Jul 09 10:11 | |
| *cubelog1 has quit (*.net *.split) | Jul 09 10:11 | |
| *TechrightsBot has quit (*.net *.split) | Jul 09 10:11 | |
| *benJIman has quit (*.net *.split) | Jul 09 10:11 | |
| *ChanServ has quit (*.net *.split) | Jul 09 10:11 | |
| *MinceR (~mincer@unaffiliated/mincer) has joined #boycottnovell | Jul 09 10:16 | |
| *pidgin_log (~roy@host109-152-24-229.range109-152.btcentralplus.com) has joined #boycottnovell | Jul 09 10:16 | |
| *libertybox_ (~liberty@host109-152-24-229.range109-152.btcentralplus.com) has joined #boycottnovell | Jul 09 10:16 | |
| *benJIman (~benji@li273-180.members.linode.com) has joined #boycottnovell | Jul 09 10:16 | |
| *TechrightsBot (~b0t@mail.copilotco.com) has joined #boycottnovell | Jul 09 10:16 | |
| *ChanServ (ChanServ@services.) has joined #boycottnovell | Jul 09 10:16 | |
| *cubelog1 (~cubeman@maxhost.org) has joined #boycottnovell | Jul 09 10:16 | |
| *banks.freenode.net gives channel operator status to MinceR ChanServ | Jul 09 10:16 | |
| **** BEGIN LOGGING AT Wed Jul 10 06:41:43 2013 | ||
Generated by irclog2html.py 2.6 by Marius Gedminas - find it at mg.pov.lt!
