From javier at candeira.com Tue Aug 5 21:39:07 2008 From: javier at candeira.com (Javier Candeira) Date: Tue Aug 5 21:39:32 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? Message-ID: <489900BB.2020905@candeira.com> For a face-tracking project I want to embed two (reasonably cheap, reasonably fast) small cameras in a project's enclosure, so they are not seen. I am thinking of the ones you see in laptop lids, but only if they are good enough for this application. As I say, I will use two of them. I want to be able to track the position of the audience's eyes (not where the gaze is aiming) in order to calculate a vantage point in real-time. The cameras will be about 1.5 m from users, who will be fairly stationary, so I need to track small changes in face position, mostly head movements and some small attitude changes (shifting weight, moving shoulders as people use their arms, etc). Here go questions and optional requirements: - is USB good enough? What other interfaces are there? firewire? - which provider/brand/technology do you recommend (or advise to stay clear of)? - what is my minimum bandwidth/resolution requirement for tracking users' eyes in real-time? - if possible, I would like the cameras to work through infra-red, so infra-red lighting can be used, and people don't realise the machine is watching them too. How would this change my hardware expenses/availability Thanks all, and enjoy your camping. Javier From mike at electricstuff.co.uk Wed Aug 6 04:49:44 2008 From: mike at electricstuff.co.uk (Mike Harrison) Date: Wed Aug 6 04:36:28 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? In-Reply-To: <489900BB.2020905@candeira.com> References: <489900BB.2020905@candeira.com> Message-ID: <9cpi94ppca38go6j5t06c4ffbi00nfq921@4ax.com> On Wed, 06 Aug 2008 11:39:07 +1000, you wrote: >For a face-tracking project I want to embed two (reasonably cheap, >reasonably fast) small cameras in a project's enclosure, so they are not >seen. I am thinking of the ones you see in laptop lids, but only if they are >good enough for this application. As I say, I will use two of them. > >I want to be able to track the position of the audience's eyes (not where >the gaze is aiming) in order to calculate a vantage point in real-time. > >The cameras will be about 1.5 m from users, who will be fairly stationary, >so I need to track small changes in face position, mostly head movements and >some small attitude changes (shifting weight, moving shoulders as people use >their arms, etc). > >Here go questions and optional requirements: > >- is USB good enough? What other interfaces are there? firewire? (Why do you need 2 cameras?) USB2.0 should be fine - abount the same speed as firewire. The main problem is likely to be finding ones with drivers that support multiple cameras. There is an industry standard called camera link, used for industrial/scientific cams, but will not be cheap due to being a specialist market. There will be camera link cams and hardware to do anything you want, given sufficient budget.... >- which provider/brand/technology do you recommend (or advise to stay clear of)? >- what is my minimum bandwidth/resolution requirement for tracking users' >eyes in real-time? >- if possible, I would like the cameras to work through infra-red, so >infra-red lighting can be used, and people don't realise the machine is >watching them too. How would this change my hardware expenses/availability CMOS and CCD sensors have high inherent sensitivity to IR, but colour cameras have an IR cut filter to prevent IR content messing up the colour balance. This filter is often in the lens assembly and can be removed, although I have also seen it integrated on the die on some cams. Look for a pale-blue element. You may also want to look at a TV capture device and an analogue camera - the latter may have more flexible lens options, but again you may have the issue of using multiple channels. There are probably some video interfaces aimed at the CCTV industry that support multiple channels, however this maybe at the expense of frame-rate. One option may be to use 2 cameras synced together (analogue cctv cams often have sync inputs, often called 'genlock'), and multiplex the outputs into one vid capture device on alternate frames. From javier at candeira.com Wed Aug 6 05:00:21 2008 From: javier at candeira.com (Javier Candeira) Date: Wed Aug 6 05:00:53 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? In-Reply-To: <9uni941ivvja5d3k10be6clj4i1sglactr@4ax.com> References: <489900BB.2020905@candeira.com> <9uni941ivvja5d3k10be6clj4i1sglactr@4ax.com> Message-ID: <48996825.9050804@candeira.com> Mike Harrison wrote: > > (Why do you need 2 cameras?) Because I am triangulating 3D position of a point (the midpoint between a person's eyes), so I need two references from two vantage points, which are the two cameras are at both ends of the screen. > USB2.0 should be fine - abount the same speed as firewire. The main problem is likely to be finding > ones with drivers that support multiple cameras. Hmm, I wonldn't mind having two different cameras from diferent manufacturers. > There is an industry standard called camera link, used for industrial/scientific cams, but will not > be cheap due to being a specialist market. There will be camera link cams and hardware to do > anything you want, given sufficient budget.... Heh, this is art, and I am self-funded, so... I was hoping to spend poor-student amounts of money. >> - if possible, I would like the cameras to work through infra-red, so >> infra-red lighting can be used, and people don't realise the machine is >> watching them too. How would this change my hardware expenses/availability > > CMOS and CCD sensors have high inherent sensitivity to IR, but colour cameras have an IR cut filter > to prevent IR content messing up the colour balance. This filter is often in the lens assembly and > can be removed, although I have also seen it integrated on the die on some cams. Look for a > pale-blue element. > > You may also want to look at a TV capture device and an analogue camera - the latter may have more > flexible lens options, but again you may have the issue of using multiple channels. There are > probably some video interfaces aimed at the CCTV industry that support multiple channels, however > this maybe at the expense of frame-rate. And money, and especially size (I really want to integrate the cameras into the frame). > One option may be to use 2 cameras synced together (analogue cctv cams often have sync inputs, often > called 'genlock'), and multiplex the outputs into one vid capture device on alternate frames. Interesting! However, seems more trouble than it's worth. J From dorkbotlist at haque.co.uk Wed Aug 6 05:16:45 2008 From: dorkbotlist at haque.co.uk (u h) Date: Wed Aug 6 05:17:06 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? In-Reply-To: <48996825.9050804@candeira.com> References: <489900BB.2020905@candeira.com> <9uni941ivvja5d3k10be6clj4i1sglactr@4ax.com> <48996825.9050804@candeira.com> Message-ID: hi javier this is a ghetto way to do it and may be totally in appropriate for you, if you're looking for something much more accurate, but i remember danny rozin had a great way of doing something like this with a single camera using his ol' TTC pro xtra for director (though it had to be callibrated). basically you capture the size of the object you are tracking at a particular moment, and its map its on-screen XY to a presumed XYZ coordinate in space. then if the object gets smaller on-screen (and following perspective moves slightly downwards on screen) you can work out how far back its gone (i.e. its Z value); if it only moves down screen but doesn't change size then its Y value has changed; and you can of course figure out the X portion of its coordinates too in the obvious way. i mention this because if you're working with something like processing and you used something like and you captured values at a particular starting point when people arrive in front of your interface, then thereafter you would simply look for *change* in XY and radius in order to infer *change* in position, rather than the more "perfect" absolute position, though you could probably approximate reasonably well the eye position. then you could probably use any easily available camera (as long as, if you want to use infrared lighting, you manage to extract its IR filter). my favourite is the old firewire isight -- but they were withdrawn due to lead content... otherwise, how about this: http://www.instructables.com/id/Making-a-Night-Vision-Webcam/ usman From dorkbotlist at haque.co.uk Wed Aug 6 05:18:11 2008 From: dorkbotlist at haque.co.uk (u h) Date: Wed Aug 6 05:18:33 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? Message-ID: oh and just found this: http://www.hoagieshouse.com/IR/ u From javier at candeira.com Wed Aug 6 05:49:39 2008 From: javier at candeira.com (Javier Candeira) Date: Wed Aug 6 05:50:01 2008 Subject: [dorkbotlondon-blabber] Cameras like the ones that fit in laptop lids? In-Reply-To: References: <489900BB.2020905@candeira.com> <9uni941ivvja5d3k10be6clj4i1sglactr@4ax.com> <48996825.9050804@candeira.com> Message-ID: <489973B3.1090908@candeira.com> Hey, Usman! Thanks a lot, and thanks a lot everybody. I feel bad constantly telling everybody "uhh, that won't work!", but I want to tell you I appreciate all your advice, you all. Y'all. > this is a ghetto way to do it and may be totally in appropriate for you, > if you're looking for something much more accurate, but i remember danny > rozin had a great way of doing something like this with a single camera > using his ol' TTC pro xtra for director (though it had to be callibrated). Right, my current masking-tape-and-baling-wire code does exactly that. I am currently cribbing from Johnny Lee's wiimote code, using a wiimote and goggles with infrared leds on the sides. However, I need more precision and the ability to track smaller head movements. When/if I get hold of a second wiimote I will be able to check that two cameras and triangulation give you enough precision, but for now I think it's a safe assumption. > basically you capture the size of the object you are tracking at a > particular moment, and its map its on-screen XY to a presumed XYZ > coordinate in space. then if the object gets smaller on-screen (and > following perspective moves slightly downwards on screen) you can work > out how far back its gone (i.e. its Z value); if it only moves down The problem is that it's easy to bias that system by turning your head, and I expect my audience to turn their heads... but I don't want the vantage point to move back like in a Hitchcock movie. > then you could probably use any easily available camera (as long as, if > you want to use infrared lighting, you manage to extract its IR filter). > my favourite is the old firewire isight -- but they were withdrawn due > to lead content... otherwise, how about this: > http://www.instructables.com/id/Making-a-Night-Vision-Webcam/ Yes, Isights are good, but again: too bulky, I am afraid. Javier From alex at slab.org Wed Aug 6 14:38:39 2008 From: alex at slab.org (alex) Date: Wed Aug 6 14:39:25 2008 Subject: [dorkbotlondon-blabber] new instruction Message-ID: <1218047919.18438.28.camel@teff> After a summer break, instructionset is back: [August 2008] Surprise According to the principle of least surprise, often cited in programming literature, software should be invisible. Write some software that only conforms to this principle for 10 seconds or so before astonishing the user in some manner. Code obfuscation is optional. http://instructionset.org/instruction/14/ You are also welcome to implement instructions from previous months. Instructionset now has its own mailing list, so this will be the last mailout to external mailing lists. You can subscribe via the website: http://instructionset.org/ RSS/Atom feeds also available. cheers From pixelsumo at gmail.com Wed Aug 6 14:59:32 2008 From: pixelsumo at gmail.com (Chris O'Shea) Date: Wed Aug 6 14:59:47 2008 Subject: [dorkbotlondon-blabber] Re: Cameras like the ones that fit in laptop lids? Message-ID: <4899F494.3010908@gmail.com> Hi Javier I had a go at face & eye tracking once, although only made a basic demo... http://tinyurl.com/5zkm5x Using opencv you can find faces in an image, then take that region and look in the top area for eyes. From there you would need to do further processing if you want to be accurate. You should check out Golans project which uses two hidden cameras that detect eyes very well (including blinking) http://www.flong.com/projects/optoisolator/ Making of pics on his flickr page: http://www.flickr.com/photos/golanlevin/sets/72157603365971672/ Best -- Chris O'Shea +44 20 8133 7462 www.chrisoshea.org www.pixelsumo.com www.thishappened.org From alex at slab.org Thu Aug 7 11:56:52 2008 From: alex at slab.org (alex) Date: Thu Aug 7 11:57:37 2008 Subject: [dorkbotlondon-blabber] slub at immersion at flea pit tonight Message-ID: <1218124613.18438.71.camel@teff> a bit short notice but we (slub) are playing in the flea pit on columbia road tonight: http://www.immersionclub.tk/ would be good to see some dorks there! cheers, alex From alex at slab.org Mon Aug 11 13:26:07 2008 From: alex at slab.org (alex) Date: Mon Aug 11 13:27:39 2008 Subject: [dorkbotlondon-blabber] dorkcamp, bentley copse, shere, 29th August - 1st September 2008 Message-ID: <1218475567.8772.49.camel@teff> Get your tents, soldering irons and knitting needles out ready for the third annual -----=================----- --++[ D O R K C A M P ]++-- -----=================----- 29th August - 1st September Bentley Copse, Shere, nr Dorking Buy your tickets here: http://dorkbotlondon.org/camp08/ Adult tickets are ?30 each, including camping and food. Bring your children for ?15 each, babies camp for free but let us know they're coming. Advance tickets only, and we have to be strict this year. Over half are sold already and we always sell out, so buy yours now! We return to the beautiful woods of Bentley Copse, where we've hired out a hall, kitchen, showerblock with flush toilets, camp circle and plenty of space for camping. Bentley copse is about an hour out of london, travel info developing here: http://dorkbotlondon.org/wiki/index.php/Travel You are encouraged to bring something dorky along to show/talk about/run a workshop about if you think others might be interested, even if it doesn't involve electricity. It's also totally volunteer-run, so you're also encouraged to help out with setting up, tidying up and running the kitchen. A taster of things happening... * Arduino tryout/troubleshoot workshop -- electronics and sensors with open hardware. Bring your own arduino board for this -- you can buy them from http://tinker.it/ukstore/ * Knitting workshop * Night sound walk * Microcontroller taster session * Movement for geeks workshop * Human live patching (like max patching but with people) * Micro scrapheap challenge (please bring some electronic junk to contribute to the scrapheap) * Musical instrument making * Fire making * Cocktail making * Evening presentations and music around the campfire including Bad Science from Ben G * The big burn! More info on the wiki: http://dorkbotlondon.org/wiki/index.php/DorkCamp08 If you have an idea for an activity, go to the "ThingsHappening" page, click edit and suggest it there! See you at the camp... http://dorkbotlondon.org/camp08/ From kat_borges at yahoo.co.uk Mon Aug 18 16:27:56 2008 From: kat_borges at yahoo.co.uk (borges kat) Date: Mon Aug 18 16:28:11 2008 Subject: [dorkbotlondon-blabber] Kat as guest in CTO Blog In-Reply-To: <1218475567.8772.49.camel@teff> Message-ID: <865051.14831.qm@web27902.mail.ukl.yahoo.com> hi guys Please check this article I wrote about The new generation of CIOs at the CTO blog of Capgemini Andy Mulholland Global CTO invited me as guest. He is a cool geek. And his blog has voted the best CIO/CTO blog by Computer weekly. http://www.capgemini.com/ctoblog/ please post comments if you want and I will reply from the blog Cheers Kat Borges Send instant messages to your online friends http://uk.messenger.yahoo.com From tom_croucher at yahoo.com Fri Aug 22 07:07:29 2008 From: tom_croucher at yahoo.com (Tom Hughes-Croucher) Date: Fri Aug 22 07:07:45 2008 Subject: [dorkbotlondon-blabber] helps with lasers and projectors Message-ID: <184080.3031.qm@web34702.mail.mud.yahoo.com> Hello Dorks, it's been a while since I've been around on the scene. I'm writing cos I need some help for a friend of mine who's staying with me. He's doing some stuff for Graffiti Research Labs in Vienna except he's in London, but his stuff is in Vienna. He's hoping someone can lend him the use of 1 Laserpen (30-100mW) 2 Projectors (2500 or more lumens) for one night in the next two weeks. He's not sure which one yet, but I'm sure he can be flexible. If anyone is able to help please drop me a mail. Thanks! Tom