Motion Capture Technician Study (Data Acquistion and PP)
01/30/2023
I have returned to volunteer my time with CSU Chico to teach and learn from the students as well as lend my expertise to them. To start off, it's a simple lesson for first timers to learn constraints to non-human assets using C3D data from Motive. I'm aiming to teach them angles, camera work, short-cuts and tricks in Mobu/Motive.
11/09/2022
I showed up early to the labs to recalibrate the stage, wand the space and set the ground plane. I try to run a demo with the current TAs to ensure they understand how important it is to set a clean, exceptional stage.
We've brought in the two teams of three people each to make sure the suits are tightened with velcro straps and got them to add on the markers to their actors. Sadly, they'd have to learn more about the spacing of each marker as I had to correct their placements. We had two actors of varying size, a small guy about 5'4" and a short woman about 5'4" whose chest was making the model on Motive look broken and hunched.
The first group with the female actress had a problem with consistently looking at the screen and causing them issues in post for countering the head and chest turning during a run through. So, me and the other TA blocked her view and got them done in the stage. Their model didn't break, and they'd only have to focus more on the TAK files that she is turning her head.
The second group with the male actor, they got in and out quickly and were efficient. That group was doing a previz of Spider-Man 3's venom suit spidey point his finger to people and dancing. They completed their shots and added in some flips and comedic shots too.
11/07/2022
I got in early to clean the stage and calibrate the system. The system crashed a few times when we ran the calculations from wanding, but third times the charm. The teams had issues with loose suits, so I taught them how to use velcro straps again.
One of the issues I saw with both groups, is that they are not thinking about the clean-up in post processing and the continuity of their scenes. I had the break it down for them that, they should run with direction and purpose to minimize the work into post. One example is that the one scene they're doing, the actor is running from a monster, and they are looking around for a window to jump out of. I pointed to them to stop in a direction they didn't look to because it made no sense that you'd be looking around from point A to B and found the window at point A.
After they understood, they completed their shots with minimal issues.
11/02/2022
Studio's dimensions: 20ft(L)x14ft(W)x10ft(H)
Composed of 16 IR Camera (Flex 13) from Optitrack
OptiWand: CW-250 (250mm) and Wand Length of 250mm
Today, each of the respective groups will have to bring their action lists and we'll be directing/assisting them in directing to get them their motive TAK files. Luckily, the studio's dimensions have not changed so the same calibration workflow from prior have worked. I got to help the current teams understand what it means to optimize their captures. Such as making sure the marker placements are rock-solid, suits are skintight, and calibrating the cameras to set the room up properly.
The first team and second team of the sessions didn't seem to understand the importance of a tight suit. I broke it down to them that a tight suit will capture the cleanest data. Having some loose clothes can mess with your capture. I grabbed two handfuls of Velcro straps and demonstrated to them on how to properly tighten things up as their actors had some snug suits, but they could be better. I had them mirror me and they were all set. The issue for this is due to the lockdown pushing lessons online exclusively, so it slipped the lesson by forgetting to mention the straps.
10/01/2022 - 11/01/2022
I have wrapped up my contract with Scanline VFX / Eyeline Studios as a Volume Capture / Motion Capture Technical Director, yet I'm glad to have worked on Black Adam, The Flash, Aquaman 2, The Batman (2022), and The Gray Man. I'm taking some time to myself and assisting the motion capture program at Chico State for the CAGD major.
I'm assisting the professor with their workflow and making comments on what can be omitted, added and rearranged in regard to the students on their respective workflows. We've completed a project involving a game animation workflow to bring the animation files Unity. With all the learn skills they have; the class is focusing on Previz's. We've had them rough out the story, storyboard it out and make a quick animatic to figure out the timing and camera angles.
10/18/2021
Class has been cancelled today, so we've extended the deadline for groups to submit their game animations via Motive, Mobu to Unity workflow. While the groups work on their projects, I've been cleaning up the data for our TA project as I was able to clean up 7 of the 9 TAK files.
This is the first time I've cleaned up a scene with two people in the scene, but they don't really make contact so it hasn't been too much work. One thing I've noticed, the hands and feet of either models have been a bit scrabbled when I scrub through the playback. I've been fixing the labeling on the markers and filling in the gaps with the "Model-Based Fill" since the "Cubic Fill" can be a hit or a miss.
The motion data is a bit jagged when zoomed in on the graph editor, so I've been hitting the Smooth Track with a default of 6 Frequency twice.
Current accumulated hours in lab: 44h
10/13/2021 - 10/16/2021
Little to go about this week, we got the room ready and calibrated for possible use. Groups already had their data captured, so the room was open for people to use the available license of Motive on the extra computer.
I spent my time helping the groups with their cleaning up in Motive and counter animations in MotionBuilder, characterizing their rigs and defining their motions as well as troubleshooting issues. As mentioned before, there is an ongoing issue with the early version of Motive we have and the latest version of MotionBuilder 2022, where we cannot export from Motive as FBX ASCIIs to the newer version of MotionBuilder. The workaround we have been using was exporting the Human capture data as FBX Binary and C3D for non-human captures i.e props.
The other TAs and I have decided to do a reVis project, and they recorded the capture data on 10/9/2021 with two actors in Mocap suits. Myself and one of the TA directing the project have split the capture clean up work load to where I clean up the first 9 takes and him the last 9.
Current accumulated hours in lab: 44h
10/11/2021
For this project, I was unable to access the lab early to get a headstart on the calibrations and wanding. Luckily, we didn't have any groups scheduled today as all had already did their initial captures. We had the room set up for any group needing access Motive as all licensed computers were in use.
Today was a rather calming day, not much done as we simply assisted people in their questions on using Motive for post-processing clean up and MotionBuilder for editing content, characterizing rigs, utilizing the animation layers and troubleshooting.
Current accumulated hours in lab: 38h
10/6/2021
We continued with starting a new session with the same group prior. I showed up early as my professor didn't have time to change the settings and try it out, so I got things ready. I got the room calibrated and wanded, then set the ground plane. But I did notice the ground plane calibrations were at 45(mm) from the previous 55(mm) offset. I thought nothing of it, until the system asked to auto-update the offset. I let it update and the new offset was 90(mm). I then noticed on the viewport the room was much bigger than before.
So, I went back to the calibration menu and saw that the defaulted wand size was 500(mm) instead of the 250(mm) we normally used for our space. I changed it back and re-wanded, then set the ground plane. I forgot the ground plane's offset default is associated with the wand size. The defaulted offset for the 250 wand was 55(mm) on the ground plane.
The system ran as smooth as butter with no lagging nor freezing. I got the group in, readied the actor and directed them with little guidance as now they have the most experience with getting started out of the other groups now.
All takes were clean with minimal clean up needed, we exported them out and re-ran them. All takes appear to have data and are good to go.
Current accumulated hours in lab: 35h
10/4/2021
Today, it was the same as usual. I showed up early to calibrate the room and wand it for the incoming group of three. However; I ran into some issues when wanding the room on the first pass. During the wanding and the computer froze. I had to hard reset the computer as 15 minutes elapsed. Then start up the program again, and finished wanding and then changed from the camera layout menu to the character creation menu. However; the computer froze again and had to hard reset another time.
I started up the computer again and while it was booting up, I assisted in applying the markers to the actor. I had someone else sign in and open the program. I used the same TTP file and CAL file that I was able to make to open the scene to skip to the actor recording.
The program and computer froze. I did another hard reset.
I tried to open a past TTP and CAL file from prior successful sessions with other groups. We were able to record 12 takes, until it froze/crashed again.
I opened the same TTP and CAL files, and we tried to record again seeing that we've improved in the number of takes. We did 4 before crashing again.
We had our professor come in to see what he can do. In short, we did a hard reset and were told to record a run and idle only. We did successfully and then pushed it by recording one more take.
It crashed again.
One thing to note, all the data is 0 kb in the folder directory until you successfully close the program. Hard resetting the program without properly closing it saves the takes with 0 kb data. So we had two takes with data and 12+ with 0 kb. We decided to cut our losses and reschedule the session while they take the ones with data.
We saved it out and have them export it out to then open it again with the program closed successfully. I copied the TAK file to the desktop and dragged it to the take folder in Motive to open it.
I got the error message that there is no data / file is corrupted, so ALL files were bad/corrupted. Our professor said to reschedule for Wednesday. We took out the computer to his office so he can rebuild it before the next session.
Current accumulated hours in lab: 32h (Three hours on troubleshooting)
10/2/2021
We completed one group's western shootout style motion data. They had 21 various shots in their shot lists and action lists. I prepped up the room with the calibrations, wanding, and prepped the markers on their actor.
One thing I noticed during our capture session and that was how little the data disappeared. I took the feedback from past sessions and thought to concentrate more of the wanding toward the center where the actor would T-pose in, start and return to to reset the model should it break.
After the first group was completed, myself and the TAs hopped into doing our project or at least a trial run. Our scene of Harakiri, the two actors duel each other, but we only had one suit. However; we shot to get a grasp on what we are doing and tried to do the entirety with one actor. Yet, we couldn't make the captures look believable. Our one actor, Raymond, was suit up and I acted as the second man, I was just not suited, and we fought each other with the PVC pipes. We didn't have much time to capture, so we postponed as we had 30 minutes to shoot before having to leave.
We agreed that we were going to be doing a two actor sessions with two suits to capture the whole scene, so this session has been extended to next week.
Current accumulated hours in lab: 29h
9/27/2021 & 9/29/2021
For these two days, I had early access to the studio, so I took the time to get it all prepped up.
I calibrated the system, wanded the space, set the ground plane and readied the suits / markers for the actors so we can save on time. Some of the groups were short one person, so I assumed the roll of directing them through their shot-list and the actions pertaining to the actor model they were trying to recreate.
Like heavy sword swinging having heft to the motion or follow-through with a running motion. During that time period, we had two groups show up on each day with either the full three person team or the lone member.
Current accumulated hours in lab: 26h
9/25/2021
We were assisting two groups today, one of three and one of one person. The first group of three had technical issues as we had to restart the computer twice. Then we had to calibrate the studio and wand the room. Due to time constraints, we had to proceed with the cameras being less than "Exceptional" and we had issues on various parts of the body snapping to other parts.
For that group we recorded a total of 30 baseball motion takes. The second group of one person came up and we recorded his 15 Geralt of Rivia actions. I fitted him to his suit, attached the markers and directed his shoot as he couldn't leave the space due to the model braking each time.
For one of our TAs, we have built out an environment for the Harakiri duel scene as well as a shot/action list. We begin recording the takes next week.
Current accumulated hours in lab: 20h
9/22/2021
I have been assisting groups with suiting up the actor as well as record their human model in addition to setting up the studio and the calibrations. With my group, we gathered our assets for the environment scene we are making for our group project for the Harakiri Duel Scene.
I'm also doing a continuation of my Joker Revis scene and I'm working on finalizing the export files for the motions with the environment in Unreal.
I have completed the counter animations for the model and pieced together for display below.
9/20/2021
We were going over the basic of the Mocap suit today as well as running a practice session of some takes. We used the 50 marker placement as per Optitrack with the model shown below as reference and myself suited up:
I have done the following:
- Downloaded my TAK files for the scenes I still needed to complete
- Retargeted them to the main model for the "Joker"
- Counter-animated the FBXs using the animation layers
- Exported them out of Mobu
- Create the environment (I accidentally deleted my previous version that was completed)
- Lighting
- Camera Sequencer
- Render Motion in UE4
9/8/2021
Current TAs and myself agreed to work on a group piece. Current idea is a ReVis of a samurai duel from Harakiri (切腹, Seppuku , 1962) is a 1962 Japanese jidaigeki drama film directed by Masaki Kobayashi. Once we get approval we'll start with the storyboarding, action list, shot list, and planning. Currently, I'm brushing up on the Optitrack system itself.
Current accumulated hours in lab: 10h
9/2/2021
We are working on a test project to get into the flow of Motive to Motion Builder creating a prop and running it through the scene. Due to certain circumstances, we are exporting the takes as C3D files from Motive to Mobu as our current version of Optitrack can no longer be supported due to the camera not being upgraded for the newer version of Motive conflicting with the newer version of Mobu '22.
Prior to the export, we are still able to clean the motion data in Motive and fill in the gaps and manualling labeling points. Once that is done, we bring the C3D data in Mobu and assign a model to it and bake on the data to then counter animate. We work on the camera shots and did a render out.
9/1/2021
Studio's dimensions: 20ft(L)x14ft(W)x10ft(H)
Composed of 16 IR Camera (Flex 13) from Optitrack
OptiWand: CW-250 (250mm) and Wave Length of 250mm
We are unable to assist with the set up with the cameras of the studios to run trials and test better setups for potential better data acquisitions. That being the case, we are needing to perfect the calibrations and wanding methods to compensate for better results. We are wanding till we get a result of around 6000-7000 per camera. However; we are experiencing some difficulties with cameras 1,6,8 and 15 so more attention is needed to wand more in their views as per viewport.
Comments
Post a Comment