WEEK 2: Motion Vectors
This week we continued learning how to efficiently cleanup a plate. We were shown various methods in which to do this.
I needed to first remove the tracker marks on the face. I started by rotoscoping each dot, using a circle and ensuring it tracked efficiently.
Next I used a merge (stencil) with the roto and the footage, then using a blur node at the right level to successfully minimise the dots. This got rid of them at first glance, with a premult after it. I again merged (over) the roto and the plate.
Then I needed to re-blur the dots, and use a shuffle and grain to soften it. I also used an edge blur on the alpha to achieve the same effect.
WEEK 3 & 4: 3D tracking
We discussed how to distort and undistort the plate we are using in Nuke, this is important for our collaborative project with the Crypt scene. We were also introduced to 3D tracking and building in Nuke, using the Scanline render node specifically.
We used an example shot to practice 3D tracking. We were advised not to track anything with reflections, such as windows or water, so we would need to mask them out of the shot. We used the CameraTracker node, inputting the relevant camera information.
For this semesters project I used the CameraTracker node, inputting the details of the camera that was used, then the number of tracks (features) we wanted. I tracked then solved the tracker, we were told an accurate result has an error usually between 0.5 and 1.
I deleted the unsolved nodes, then recalculated, refining the accuracy of the pass.
I added an origin and ground plane with the points, this ensured the tracks were organised and correctly placed. Then I created cards for for the walls and ground, and merged them over the scene.
Once completed I exported the trackers and camera to Maya.
In class we then discussed to use of PointCloud and ModelBuilder on our example plate.
In week 4 we looked at Projections, first establishing the difference between using a textured card and projecting onto shot. We were taught the difference between cameras whilst using a projection, and the method of using a ‘project3D’ node with a patch.
The use of a framehold node was also explained, firstly when used on a camera, then when used on a scanline render.
WEEK 5 & 6:
We continued looking at projections and the problems that may arise such as stretching, doubling and resolution issues. We also went through the various different projections that you can do.
This week I also continued working on my Crypt shot, I tracked the front wall in Nuke, using one of the projection methods we were taught.
I rotoscoped the opening in the wall on frame 0 as it had the whole area in shot, I then used a framehold on it and projected it onto a card within the scene using Project3D. When I did the roto, I needed to apply an invert node, so that the correct part of the shot became the alpha. I also added an edge blue to soften the edges, for a more realistic roto.
Week 8 & 9: Green Screen
We started learning how to manage working with green screens in Nuke, firstly however we were shown how you can manipulate an image using the colour controls. Specifically we were shown how to use Keyer and Colourspace nodes.
For the homework, I edited the background using the luminance key in alpha, then blurred and channel merged it over the plate, following a method we were taught in class. I graded this to create a pinkish glow, this is because the foreground is red tinged so I wanted them to match.
I blurred the entire background slightly. Then I used a keylight to get an alpha for our front image, denoised it and then added an edgeblur.
Once complete, I merged them over each other.
WEEK 10:
To create the final shot, I used my previous roto and merged it over my render of the stationary steam engine. I had already created realistic lighting in Maya, and input all the correct AOV’s.
I followed our tutors script to correctly grade and colour-correct my sequence, I needed to seperate the AOV channels, I mainly used the diffuse_direct, diffuse_indirect and specular_indirect. I graded and coloured them individually, to get accurate lighting. Most importantly, I wanted to get the black points of my render and the shot to match.
When happy with the grading, I applied a write node, and rendered out the overall sequence.
Then merged the roto over my graded footage to create the final shot.
I managed to remove the markers on the floor, using 4 tracker nodes and rotoing them. Then I blurred the roto until the trackers were gone, I pre-multiplied then rendered out the final footage. I am having trouble with Nuke as it is not following the tracks correctly, I have to move them on each individual frame, which is taking too long.
An issue that I have been having is that the track wobbles a lot, when the scene is in Maya it is fine, however when I render it out and import to Nuke the track shakes again
I have tried so many different ways to fix this but keep coming to the same issue, I don’t have time to re-do the camera track from scratch. Ideally I would do that and start the process again, I will try this after the submission. Overall however, I am not happy with this which is unfortunate as I have completed the model with animation and textures, and graded it to the scene.
I am attempting to fix this problem in Nuke, I started by tracking 4 points on the machine, then changed the node to ‘remove jitter‘.
So it didn’t really work. The shot is still very unstable, I really can’t understand why when the track works perfectly fine in Maya. I will need to speak to a tutor regarding this issue.
I have found out what the issue was, the frame rate of the footage and the machine animation were not the same. I will render out a new sequence to input into my Nuke script, which should be stable.