Tag Archives: matchmove

Making Of : Philips Carousel

I think this is a really well done commercial. I wonder how they did it for a while (bullet time ? tracking ? green screen? ..) it seems they used a mechanical dolly controller which could reproduce the exact same path several time, and some stunt guy who were good at freezing body :). they shoot it one time with the stunt guys and another time without. Really cool !

MatchMove with Syntheyes a Nikon D90’s plate

Those days I’m working on Reels, at least one for the moment, a matchmove one. So far I’m putting together some pieces I did at my Fxphd’s classes, and also some of my shoot.

This one I did with my Nikon D90 which is none for a rolling shutter issue (jelly video look) when shooting video.

so those were the challenge for this plate :

  • shoot from the back of a car on small road
  • hand held
  • 24fps
  • 1280×720
  • 33mm

To be honest, the hardest part were dealing with the Rolling Shutter, this one is a real b**** ! it doesn’t show to much on the video, but when you look closer the image keep getting distort all the time which give a high error to the tracking data. Plus on top of that you have fast movements which give the “debug tracker” part a real nightmare. It’s not perfect, but I do think it’s looking very well.

you can watch the result in HD on the vimeo website


[MatchMove] : small road with Nikon D90 from François Tarlier on Vimeo.

Part of the shot I did track for my reel.
The particularity of this one, is the hand held kind of shot with the “Rolling shutter” issue of the D90.
It looks easy with all the features but it was a really informative one to track. I learned a lot with this shot, like do not try to get it perfect! at some point, I got a lower error (right now it’s 0.8hpix) about 0.4, but it wasn’t looking as good as this one. I guess it good be much better at some point but so far I think it works pretty well right now.
It has been track with Syntheyes, Blender for the 3d grid, and After FX for the compo/roto

Fake Running Crowd on Vimeo


Fake Running Crowd from François Tarlier on Vimeo.

This class (which was teach by Gareth Edwards)was really amazing, I highly recommend it !! I didn’t learn so much about AfterFX but mostly about how to get the job done in a small amount of time, and this goes through “organization” !! Keeping your workflow as clean as possible, taking time to renaming everything (cuz god it could get really messy in AE),…

My result is far from being perfect, I can see so many things I could change (the way they are running, the dust, the lighting, the shadows, … ). but so far, I think it’s quite good and I have so much other things to do that I cannot work any more on this piece! I’ll probably make a small “making of” video later on, just to show all the passes (shadows, AO, Dust, lighting,..) were done in After Effects, only the 3D tracking was done with Syntheyes, but so far it was a super easy one !

What I learned with this project :

  • Many AE tricks : can’t tell all of them, but I definitely think I’m more efficient now with AE. I’ve done many mistake and encounter many problems that did teach me a lot 🙂
  • Organization : I won’t say it enough, it might sounds silly but when you look at all the job that Gareth has done on “Heroes and Villains: Attila the Hun” (VFX breakdown here) in such a small amount of time, you have to say he had a pretty good workflow. So far that’s the reason I took this class and I don’t regret it
  • Faking : That’s mostly what I like in compositing, is that you can fake so many thing which usually with a good amout of blur you won’t see the trick 😉 . But even with that, we usually go for the hard way (3d, particle system, IA, ..). As I said, the result is far to be perfect, but I guess you have to keep in mind that the quality of your work could depends of the target you have and of course the budget (cf the production triangle). In this case the target is TV Shows, so I guess it should be “Fast” and probably “Cheap”, which leave “Good” a bit behind. Of course it should be as good as possible and that’s why I think knowing how far you can go with faking is a really good thing, and so far that’s also something I did understand more in this class.

Demo of Adobe Flash Face Recognition using the openCV port to AS3

Click on the picture to try it
Click on the picture to try it

see also the previous posts here : http://www.francois-tarlier.com/blog/?tag=marilena

Source code here : http://www.francois-tarlier.com/blog/?p=408


[iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=0596516134 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=3836499320 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=1934356034 120 240]

Playing with Marilena

this is my first try with OpenCV port to Adobe Flash (Marilena). Meaning the entire face recognition has been done with Flash/ActionScript

For you who doesn’t know yet about Marliena, please have a look at my previous post “MariLena : OpenCV port to ActionScript 3 (AS3 / Flash)”

Source code here : http://www.francois-tarlier.com/blog/?p=408


Marilena : OpenCV for Flash on Vimeo.

Playing around with the port of OpenCV face recognition to ActionScript 3. In this test I did use Mario Klingemann’s optimized version of Marilena (code name of the port).

Would be great to have this kind of feature on Glasses seller website 🙂

next step try to get the Roll rotation 🙂

Mario Klingemann’s Marlena version could be download here : http://www.quasimondo.com/archives/000687.php#000687

see my post about Marilena : http://www.francois-tarlier.com/blog/?p=235

[iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=0596516134 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=3836499320 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=1934356034 120 240]

Reblog this post [with Zemanta]

Deface : AS3 Face Detection BASED on OpenCV

As my previous post about MariLena (openCV port to AS3), I made more research about it and found this project Deface, which is pretty much the same as MariLena, except that this is not a port of openCV, as the developer says “it’s written from scratch”. so far it’s using Haar cascade file as OpenCV.

I haven’t tried it that much, here is an example : LiveExample
It seems to be a bit slower that MariLena, but I wouldn’t be sure at all

MariLena : OpenCV port to ActionScript 3 (AS3 / Flash)

I dreamed about it, they finally did it ^^.

A crazy guy named Ohtsuka Masakazu have ported the face detection part of OpenCV to AS3 and added the source code named Marilena to the Spark Project a few weeks ago (can’t believe I haven’t notice before today)

Mario Klingemann from Quasimondo.com already made some optimization (see his blog)

I made some test today with a stupid texture, starting from Mario’s webcam example and that’s just fun !!! Well yes, it’s still slow (15 fps on a good computer), but this is just the beginning as I believe that a lot of optimization can still be done, and so what anyway ? It’s flash, it’s everywhere, nothing to install, the fun is there without downloading anything. I really believe in it ! Now with this and FLARToolKit (ARToolKit port to AS3), augmented reality and computer vision is really getting into Flash for sure, and it’s going to give to the community some great opportunity !! I’m looking forward about it

Please, have a test at Mario’s example here

and take also a look at what they did at Squidder.com


Realtime Face Detection in Flash from squidder on Vimeo.

For more info, check out http://www.squidder.com/2009/02/26/realtime-face-detection-in-flash/

[iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=0596516134 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=3836499320 120 240] [iframe http://rcm.amazon.com/e/cm?lt1=_blank&bc1=FFFFFF&IS2=1&bg1=FFFFFF&fc1=3F3636&lc1=E6005C&t=francoi-20&o=1&p=8&l=as1&m=amazon&f=ifr&asins=1934356034 120 240]

Reblog this post [with Zemanta]

[My Work/Job] YouUp : Finally Out (kind of)

YouUp is a Ubisoft’s project, which I’ve been working on for the past 3 years. First as CG Artist / R&D with the real-time tracking technology and the internal 3d engine, also as Modeler/Mapping for the first assets we have made and finally as a TD Artist (designing features, tools, workflow, techniques, assets production pipeline, training graphic team, and so on).

It’s been a long way since then, this is the really beginning of this kind of technology, a lot more will come out of our work beside youup (so I hope).

For now you can download YouUp Windows Live Messenger version, which is a special version we brought with a Microsoft paternship. Download it here : http://www.ilovemessenger.fr/youup

You can also watch this interview (it’s in french, but you can only watch it, what they say is not that important, I appear in the end, that’s the important part 😀 )

You Up: la nouveauté révolutionnaire sur MSN – Nouvel Obs
Ubisoft propose de rendre MSN plus ludique en transformant son visage en temps réel pendant les t’chats. “YouUp” est téléchargeable gratuitement sur www.ilovemessenger.fr/YouUp.

FxPhd : Jan09 first 2 weeks bilan

fxphd Here we are, already the 3rd week! So now it’s done, for sure I’m taking the following 4 classes for this term:

  • MSV101 – Introduction to Massive
  • AFX301 – A Guerrilla Filmmakers Guide to After Effects
  • PFT202 – PFTrack in Production
  • FUS101 – Introduction to Fusion

MSV101

This class is really nice, but in the other hand, massive looks like the most complicated software I ever “had to” learn. But it’s really interesting and I can’t wait to see what is going to come out of it… Continue reading FxPhd : Jan09 first 2 weeks bilan

Libmv : making blender a tracking/matchmove software

libmv is a structure from motion library, which plans to one day take raw video footage or photographs, and produce full camera calibration information and dense 3D models.

The first targeted application for libmv is to integrate with Blender to make the first open source matchmoving solution

Roadmap – libmv – Release plans and goals for libmv. – Google Code.

I really looking forward for this project, since it will bring blender to a next level of vfx tool and make it the free complete opensource matchmove application available on “the market” 🙂

Check out the following video of what they got so far : Continue reading Libmv : making blender a tracking/matchmove software