Fusion Question For Experts On 3D Camera Tracking Exercise??

Can someone at BMD (or an expert on this forum) help sort out the last question I had at the end of Lesson 12 in the VFX Guide please?

To keep this really simple, organized, and to the point, here is a Dropbox folder containing the .dra, a separate export of my .comp from the exercise, and the VFX Guide .pdf : https://www.dropbox.com/scl/fo/nm5b3qta ... hmfe2&dl=0

The question/issue involves values on Camera3D1 after Export from the CameraTracker, after creating the mattes for the areas I wanted to avoid.

The BMD completed one is in the Lesson 12 Completed timeline. The one I worked on is under Lesson 12 Start. In the comp I tried to keep things clear. On the left side is a series of nodes containing a set of mattes that was not keyframed, so it drifts off the subjects. THIS is exactly what I found BMD did in the Completed timeline. But it's not what I originally did....

Because previously we had keyframed everything throughout the .comp, I went ahead and rotoscoped everything like usual, and when I got done and performed the Export, things were very weird. The flag seemed to move and never looked tacked down to the rocks. Things were generally off. Looking closer, the Far value under the Camera3D1 node was low - as in, sixty-six.

But when I didn't keyframe at all the way that the Completed timeline from BMD has no keyframe on the mattes....the Far value was 1078 and everything looked much better.

Digging deeper into the Reference Manual born out of complete and utter frustration and despair with the holes and omissions throughout the VFX Guide, at least I did learn a little more about the Camera3D near/far values.

But I'm still very unclear on what is shifting the value that much just because I adjusted the mattes throughout the render range vs. a sloppy few mattes that are not but somehow....that's what in the BMD completed timeline and what seems to work better in this case.

And I'm also not entirely clear on why if I copy the ImagePlane3D node from the completed timeline into my .comp in the Lesson 12-Start timeline, things don't line up with the transform values and the banner doesn't look as good as it does in the Completed timeline.....my suspicion is it may have to do with the Ground plane settings.

My computer is slow and laggy to work with the practice files anyhow in this 3D exercise and I'm not about to buy a new computer in the immediate future, and whether or not I even need to do 3D Camera Tracking at all for guitar videos seems unlikely.

But after the process I just went through for weeks (and it was excruciating) chronicled in a really long thread containing all the typos and omissions throughout the entire VFX Guide (posted on the Resolve side of the forum), I'm not about to let the last Practice Exercise defeat me.

So what's going on here? What did I miss? Or what else was not stated in the guide that affected the outcome?

Thanks,

Sean

Sean, if I have time tonight, I'll check out the chap 12 lesson and run through it and see if I can spot anything for you.

Thank you Ernest, whenever you get a moment is soon enough & just keep me posted. Figuring out what's going on with the CameraTracker/mask combination on that last Practice Exercise is the only thing standing between me and finishing the course so whatever advice you have is sincerely appreciated!

Sean,

I wasn't able to download your Dropbox file, it looks like the URL is not complete. Regardless, I ran through the exercise.

For the Flag exercise, I think the issue might be with the mattes you used. The chapter states that the 3D tracker (when using mattes) will track what is white and disregard what is black.

So, we want to make sure your mattes are setup correctly. So, step 2 of the exercise:

2. Create mattes for the areas you want to avoid (pirates, swashbuckler and ocean).

If you create mattes around these items and connect the mattes together, then, they create a black and white matte. However, unless we change some things in the inspector for each of the mattes, you will be tracking the wrong things.

If you connect your mattes together and view them in the viewer, you will probably see that the pirates, swashbuckler and ocean mattes are white and the rest of the area is black. We need to change that.

So, I have my mattes as ocean, swashbuckler, front pirate and back pirate. So first, I need to Invert my ocean matte. We do this because we want to exclude the ocean. Then, I connect my ocean matte to the swashbuckler matte. For the swashbuckler matte, I need to change the Paint Mode to subtract. Then, connect to the front pirate matte and choose Subtract. Then, connect to back pirate matte and Subtract.

Now, choose the back pirate matte and view it. You should see a reverse of when you first viewed the mattes. Now, when you plug this into the 3D tracker node, it will track the white areas and disregard the black areas (the ocean, swashbuckler and pirates).

Let me know if this helps. The manual does not go into depth in that chapter about mattes enough to allow you to know the different options for the mattes. However, when creating mattes, understand that what you are wanting (in regards to 3D track mattes) is to create the white areas for it.

Sean,

I wasn't able to download your Dropbox file, it looks like the URL is not complete. Regardless, I ran through the exercise.

Thanks again Ernest. The URL was my mistake. I copied and pasted that post from a previous posting elsewhere on the forum and when I did, the full link didn't cross-pollinate. Here is the real link so that you can see what was going on in my comp: https://www.dropbox.com/scl/fo/nm5b3qta ... hmfe2&dl=0

It's great to get a refresher on the mattes....I just checked though, and what you suggested is exactly what I did for the mattes (although I think I probably didn't mask the swashbuckler, if I recall, because -- despite the instructions -- I found out the BMD completed comp didn't either, and I started trying to figure out what they did that I didn't that could account for the issues I was running into).

Things seem to be "okay" when the mattes are not adjusted over the render range, but as soon as I do, things get much more bizarre. Again, I noticed that the Far value on the Camera3D node after Exporting the CameraTracker (thus generating those five nodes referenced in the guide) is dramatically different depending on whether the matte is keyframed throughout the region or not.

If you have a second to look at the comp via the Dropbox link that should now work, I'd be interested to hear what you think, and thanks again as always.

Sean,

Dropbox wants a login to download file. How about this...drag and select all your nodes in the node view. Ctrl-C or Command-C (depending on Win or Mac) and then paste into this chat. It will appear as code, but, it's only code for your comp. That way, I can copy your code and paste it into Fusion and see your node setup.

I have the footage; so, I can just attach where needed.

Sean,

Dropbox wants a login to download file. How about this...drag and select all your nodes in the node view. Ctrl-C or Command-C (depending on Win or Mac) and then paste into this chat. It will appear as code, but, it's only code for your comp. That way, I can copy your code and paste it into Fusion and see your node setup.

I have the footage; so, I can just attach where needed.

Hi Ernest,

I tried to copy the code a few days ago and re-tried after your request, but it won't paste and I think it's probably because the comp has too much going on in it. Even trying to paste it into a Pages document just halted on my Mac.

Instead of messing around with Dropbox then, I just uploaded a .zip file of the comp only on a throwaway page I just put together on my website (and promptly unlisted the url from search so hopefully this page never populates with Fretboard Playground search results! I'll delete the page once you get the file).

You can just go here instead and download the .zip file via the button link. Hope this helps and thanks again! https://www.fretboardplayground.com/forums

Sorry if you have to go through the hassle of closing the cookies window --- I was told I was required by law to have that on the main site even though I can't stand the notification pop-up myself!

Sean

Sean,

I was able to download your .zip file and have a look at it. I'm not sure what issue prevented you from copy/pasting node view, but nonetheless, I am able to view the comp.

I wouldn't fret the camera Far value. It just represents a coordinate system. Say we are standing at one end of a football (soccer) field. We know that it's a 100 meters to the other end (made up of 1 meter per unit of increment). So, we will have 100 units of measurement.

Now, say we represent that 100 units as 10 units. It will be the same length, but our coordinate system will have a different measuring unit. When I did the camera solve, my far point was 24.3.

So, as you can see, each camera solve can be different. You probably won't get 2 camera solves the exact same. We can try, but, the 3D tracker is trying to "solve" a real world camera move in 3D space. It doesn't know how far things are from each other. There is a spot in the 3D tracker that allows you to set a "Scale" unit between points that will give you a more measured "unit".

Say, for instance, we are tracking a shot of moving across the side of a house and there is a window in the wall. We know the window is 1 meter tall. When using the Scale function in the Export tab of the 3D tracker, we can tell the 3D track that the Scale "unit" between a point on the top of the window and a point at the bottom is 1 unit. We know it represents 1 meter, but the 3D tracker takes that unit measurement and will adjust its coordinate system to accommodate all points based on our Scale unit.

I looked at both of the examples in your file. Your 2nd example seems ok. I think that you are "seeing" the flag seems out of place. However, it's just positioned in a spot that is visually showing you more parallax than what you might expect. I repositioned the image plane in a spot where the cliff edge and beach meet and it appears more correct. The solve you got on the 2nd example is really good at solve error of .1734. So, we know that the solve is really good. Now, we (as compositors) have to position things in this 3D tracked "space" based on what is visually correct. If you are wanting a more precise placement of items, there is a way to select a certain point in the 3D track and have an item "placed" there and use the point's transform as the anchor.

But, that gets into more advanced, and I'm not sure a lot of this stuff will benefit your needs for the type of videos you are wanting to create. However, if you want to place yourself playing guitar on a beach, it might be of value. haha

If you are wanting a more in depth tutorial on 3D tracking, I suggest this video -

Thank you again Ernest, I sincerely appreciate your time, response, and for taking a look at that.

Hearing about the camera Far value variance helps to clear things up for me, as did messing around more with the comp after reading your reply. Yes when I reposition the image where you did/suggested, that does appear more correct to my eyes.

I'll definitely check out that video you linked at the bottom of your reply when I get an hour or so to really focus on it.

Much gratitude and your reply helps a lot!