HVX200 3D Tracking test...
- Thread starter Jim Arthurs
- Start date
Taylor Moore
Veteran
Hey guys really awesome work...
I used to use the 2D/hack 3D tracker in Flame/Inferno; saved my ass or shall I say the director's ass quite a few times....
I used to use the 2D/hack 3D tracker in Flame/Inferno; saved my ass or shall I say the director's ass quite a few times....
Shiloh Arts
Well-known member
Jim,
This may sound silly, but what do you mean by Matte (not really sure what that is)? I know call me a newbie:huh: :laugh:
This may sound silly, but what do you mean by Matte (not really sure what that is)? I know call me a newbie:huh: :laugh:
danm
Active member
A matte is a user-defined area of exclusion.
I've just ordered my SG35 35mm adapter for my DVX. I'm planning on doing a bit of 3d tracking. Am I going to have a problem tracking footage shot with a shallow depth of field? How does a program like Icarus handle tracking bg data which is out of focus?
I've just ordered my SG35 35mm adapter for my DVX. I'm planning on doing a bit of 3d tracking. Am I going to have a problem tracking footage shot with a shallow depth of field? How does a program like Icarus handle tracking bg data which is out of focus?
Greggl
Well-known member
Its going to be a nightmare. Optically, you've also got a mess on your hands.
For effects shots, you're going to want to minimise your attempt at a shallow
DOF and you're also going to want to shoot a distortion chart to map the
barrel distortion. Removing as much of the optical 'bowing' in the image
before you track, will give you better results.
Ideally, you'll place out tracking markers that can be digitally painted out
after the shoot. The type of marker really depends on the shot and which
tracking software you're using. Some like trianges, some like spheres/circles
and others really 'feed' on crosses.
Whats the shot?
For effects shots, you're going to want to minimise your attempt at a shallow
DOF and you're also going to want to shoot a distortion chart to map the
barrel distortion. Removing as much of the optical 'bowing' in the image
before you track, will give you better results.
Ideally, you'll place out tracking markers that can be digitally painted out
after the shoot. The type of marker really depends on the shot and which
tracking software you're using. Some like trianges, some like spheres/circles
and others really 'feed' on crosses.
Whats the shot?
Greggl
Well-known member
Without ground contact, it'll be a lot easier. You've still got some issues that
are going to be hard to deal with. Tracking shots that are perpendicular to
the direction of movement have very little parallax and trackable elements
are going to come on screen, travel across the screen and then go offscreen.
With little parallax, you ususally get bumps in the solution when those
tracks 'turn on' or 'turn off'. This is for two reasons, going offscreen is
the obvious one, but the closer to the edges of the frame they get, the
more they are affected by barrel distortion.
If your bg is mostly sky or seagrass, you'll have very little that is trackable
back there. You never want to track anything thats moving for your
camera solution. So be on the lookout for trackers that latch onto T
junctions where say, two blades of grass cross, and it tracks the junction
where they meet. Waving grasses, etc will also throw your track off.
Your best bet may end up hand tracking the whole thing. Smooth out the
camera jitter with a standard 2d track in you compositing software. Then
either run that through your tracking software or build a simple wireframe
ground plane in your 3d app and animate the camera by hand. In the final
comp, you can put the real world camera jitter back in, but not having it for
the track will make the tracking/3d much easier.
are going to be hard to deal with. Tracking shots that are perpendicular to
the direction of movement have very little parallax and trackable elements
are going to come on screen, travel across the screen and then go offscreen.
With little parallax, you ususally get bumps in the solution when those
tracks 'turn on' or 'turn off'. This is for two reasons, going offscreen is
the obvious one, but the closer to the edges of the frame they get, the
more they are affected by barrel distortion.
If your bg is mostly sky or seagrass, you'll have very little that is trackable
back there. You never want to track anything thats moving for your
camera solution. So be on the lookout for trackers that latch onto T
junctions where say, two blades of grass cross, and it tracks the junction
where they meet. Waving grasses, etc will also throw your track off.
Your best bet may end up hand tracking the whole thing. Smooth out the
camera jitter with a standard 2d track in you compositing software. Then
either run that through your tracking software or build a simple wireframe
ground plane in your 3d app and animate the camera by hand. In the final
comp, you can put the real world camera jitter back in, but not having it for
the track will make the tracking/3d much easier.
help
help
hello
there's something worng with my track...auto track shows good result but when i render from max it shows diffrent result....teapot is still shaking .
http://qlove.co.kr/1/333.JPG
http://qlove.co.kr/1/1111.JPG
http://qlove.co.kr/1/22223.JPG
http://qlove.co.kr/1/44444.JPG
http://qlove.co.kr/1/1.avi
help
hello
there's something worng with my track...auto track shows good result but when i render from max it shows diffrent result....teapot is still shaking .
http://qlove.co.kr/1/333.JPG
http://qlove.co.kr/1/1111.JPG
http://qlove.co.kr/1/22223.JPG
http://qlove.co.kr/1/44444.JPG
http://qlove.co.kr/1/1.avi
Jim Arthurs
Well-known member
Hi Jung!
I can tell one major problem, and that is your camera isn't doing much more than a pan/tilt. This is very difficult for Icarus/PFtrack because it needs to see some dolly or tracking motion to really work well.
Try using setting up the camera to do a 2D track (in Icarus it was originally called Camera Motion/Pan/Tilt/Zoom) and compare the results. There will still likely be some float, and this is caused by yet another issue...
... I *think* you're on a tripod (or a very steady handheld shot) and since the nodal point of the camera isn't exactly on the pan/tilt axis (and understanding a "nodal point" could open up a whole conversation on its own) it truly is a tiny, tiny dolly and track move instead of a pure pan/tilt like the software expects.
This same set up, pointed at a distant landscape would have better results, because the relative size distance between your camera and subject would be much greater than now, both compared to the distance the camera is off nodal. Whew!
What I'd do for the sake of getting a good success on a 3D track is this... hand hold your camera, make sure the image stabilization is off (if it's on for this shot you tried, that's yet another factor in causing a failure) and shoot the scene again, only this time move along the length of your shot, while panning and tilting. Or move toward it from a bit of distance. This will give you MUCH better results!
Give that a try and share back with us!
Jim Arthurs
I can tell one major problem, and that is your camera isn't doing much more than a pan/tilt. This is very difficult for Icarus/PFtrack because it needs to see some dolly or tracking motion to really work well.
Try using setting up the camera to do a 2D track (in Icarus it was originally called Camera Motion/Pan/Tilt/Zoom) and compare the results. There will still likely be some float, and this is caused by yet another issue...
... I *think* you're on a tripod (or a very steady handheld shot) and since the nodal point of the camera isn't exactly on the pan/tilt axis (and understanding a "nodal point" could open up a whole conversation on its own) it truly is a tiny, tiny dolly and track move instead of a pure pan/tilt like the software expects.
This same set up, pointed at a distant landscape would have better results, because the relative size distance between your camera and subject would be much greater than now, both compared to the distance the camera is off nodal. Whew!
What I'd do for the sake of getting a good success on a 3D track is this... hand hold your camera, make sure the image stabilization is off (if it's on for this shot you tried, that's yet another factor in causing a failure) and shoot the scene again, only this time move along the length of your shot, while panning and tilting. Or move toward it from a bit of distance. This will give you MUCH better results!
Give that a try and share back with us!
Jim Arthurs
..
..
i just exported to lightwave and it seems to work ok....i'm not lightwave user so i'm still strugling with 3ds max.
http://qlove.co.kr/1/9.mov
..
i just exported to lightwave and it seems to work ok....i'm not lightwave user so i'm still strugling with 3ds max.
http://qlove.co.kr/1/9.mov
Jim Arthurs
Well-known member
jung kyu said:
That does seem to line up better... which could indicate that there is some sort of aspect ratio issue with the background image in Max, or some sort of frame-rate difference between Max's frame rate and your clip's frame-rate. I saw a "pop" or two in the motion of the Max clip... sometimes that happens when forcing a clip to a different frame rate.
As to my earlier post... I think PFTrack has advanced quite a bit since it was Icarus... no way you'd have success with a 3D track on something that was basically a 2D pan/tilt...
Jim Arthurs
Edited: Oh, and you need to set your XYZ coordinates up... your ground plane should lie in the same plane as the table top as per this image you sent http://qlove.co.kr/1/1111.JPG so check out the user info on how to do this... it will make things easier to work with in the 3D package...
Jim Arthurs
Well-known member
Greggl said:
Gosh I couldn't agree more. You've got all the optical issues of the HVX lens added to all the optical issues of the 35mm lense (still lenses worse than good cine lenses, but still) and then any alignment issues with the camera and the ground glass. The end result is a distortion solution set way more complicated than it needs to be. I'm not saying it won't work, because I've seen stuff work that shouldn't, and stuff that should work fail, but man...
Gregg, did you ever work with Boyd Shermis? I see you did some tracking on "Gone in 60 seconds"...
Jim Arthurs
yes
yes
i finally did it..
http://qlove.co.kr/1/yes.mov
it was my mistake..i rendered backgound movie file with 3d directly in 3ds max...i thought it would work that way..thanks jim
reminding me about XYZ coordinates
yes
i finally did it..
http://qlove.co.kr/1/yes.mov
it was my mistake..i rendered backgound movie file with 3d directly in 3ds max...i thought it would work that way..thanks jim
reminding me about XYZ coordinates
new hd camera is comming out
new hd camera is comming out
cheack this out...i think this camera is better than HVX.. it records 5mega pix 720p that;s almost film look.
http://www.sanyodigital.com/HD1/features.html
http://www.sanyodigital.com/interactive_demos.html
i might sell my hc1..to buy this..hd1 bit rate is 9mb..it's enough bit rate for 720p and mpeg4 codec is better than mepg2.. sounds too good for $800
new hd camera is comming out
cheack this out...i think this camera is better than HVX.. it records 5mega pix 720p that;s almost film look.
http://www.sanyodigital.com/HD1/features.html
http://www.sanyodigital.com/interactive_demos.html
i might sell my hc1..to buy this..hd1 bit rate is 9mb..it's enough bit rate for 720p and mpeg4 codec is better than mepg2.. sounds too good for $800
Ray Smithers
Member
VERY nice work Jim.
I looks like you did not have to "dirty" up the CG very much for a match shot.
I almost cry when I have to crap up great CG when working with SD DV with poor color space.
Am I the first one to say "Can't wait for my HVX-200"?
Thanks for the post Jim.
Ray
I looks like you did not have to "dirty" up the CG very much for a match shot.
I almost cry when I have to crap up great CG when working with SD DV with poor color space.
Am I the first one to say "Can't wait for my HVX-200"?
Thanks for the post Jim.
Ray
Jim Arthurs
Well-known member
Ray Smithers said:
Thanks Ray... no, I didn't have to do much more than add some edge enhancement to simulate the detail level that Kaku used and blur it back a bit.
When I match elements to DV 4:1:1, I often change the color space from RGB to YIQ, and literally destroy the chroma sharpness of CG by scaling it to 25% wide and then stretching it back to full screen to fit the luma. That's cruel and unusual punishment for innocent CGI...
Jim Arthurs
Anhar_Miah
Veteran
Hey Jim, thats some great talent you got there !
I vote you to start a new Article/thread, basically a "Dummies guide to camera tracking" tutorial right here on DVxuser ! that would be a great help to others here no doubt.
Anhar
(I'm learning 3d max)
P.S when you camera track, how does one tackle the problem of reflection and shadow? and what about HDRI type renders (can you use them in a camera track?)
I vote you to start a new Article/thread, basically a "Dummies guide to camera tracking" tutorial right here on DVxuser ! that would be a great help to others here no doubt.
Anhar
(I'm learning 3d max)
P.S when you camera track, how does one tackle the problem of reflection and shadow? and what about HDRI type renders (can you use them in a camera track?)
Jim Arthurs
Well-known member
Anhar Miah said:
Thanks Anhar... a topic on tracking is a good idea, and since the tools are becoming very affordable tracking should be part of everyone's CG toolkit.
I think when my HVX ships I'll shoot a couple different styles of background plates and post the footage so everyone can experiment with whatever tracking software you are using.
As to reflection and shadows... if your automatic trackers are failing because of these, it's time to hand track a number of features and use those. Most of the time I find the auto trackers will self cull out track points that don't make sense in the whole of the shot. Any tracks you do by hand are treated as "absolutes" and the rest are suggestions and refinements, if that makes sense.
If you're talking about using CG elements rendered with HDRI lighting tricks, or rendering out CG elements at higher than 8 bit depth, sure, no problem. Tracking is only impacting the camera path, not the actual renders techniques in any way.
Jim Arthurs