Research Artefact 6: Mocap vs Key-Framing

This artefact takes the motion created in artefact 5 and discusses the benefits and negatives compared with key-framing the same motion. Both techniques are very time consuming but each have their own plus points. The recorded time for creating the 15 seconds of motion capture data that replicates the same motion as the key-framed animation, minus facial and finger animation as well as the tests sessions was: Half an hour set up, 1 hour capturing (several takes), 3 hours cleaning the data, 1 hour plotting the motion to an actor and character, overall time taken 5 hours 30 minuets. The time taken to key-frame the same motion minus facial and finger animation was 4 hours. This was a surprising result as previous expectations were of the key-framing technique requiring more time. Due to this outcome a second motion capture session was performed to cover 60 seconds of motion. The recorded time for this process was: half an hour set up, 1 hour capturing (several takes), 3 and a half hours cleaning the data, 1 hour plotting the motion to an actor and character, overall time taken was 6 hours. The time taken to key-frame 60 seconds of similar animation is roughly 16 hours. Research on the topic concludes that to book a commercial motion capture lab session costs around £3000 for the day, not taking into account extra cost such as actor’s fee. Based on research from Google Answers, available at: http://answers.google.com/answers/threadview/id/784826.html, the cost for an animator to key-frame this motion is $6000 or £4129.42 ($100 per second of finished animation). The positives concluded from the artefact show that extreme accuracy can be achieved when key-framing a motion, using motion capture is a cheaper and faster alternative but doesn’t allow for the same accuracy to be achieved.

Research Artefact 5: Mocap for Character

This artefact explores the process of creating motion and plotting it to a character to be rendered out in 3D Studio Max. The motion decided on was chosen for it’s similarity to an existing animated sequence previously created using key-framing. The animated scene is of a girl walking into a room, picking up a ball and looking around, which lasts around 15 seconds. Markers were set up in the lab to give rough positions for parts of the motion, these include which directions to walk and ball placement. The reflective marker and lab set up was arranged in the same way as the most effective capture test previously conducted for artefact four. This meant less time would be spent on getting a good capture and more could be spent on getting the right motion for the character. Several attempts were made to get the right motion to give the same curious actions as key-framed character. Now that a successful capture was taken, the data was then cleaned and converted to a C3D file ready to import to Motion Builder. The data still had imperfections so some time was spent fixing this in Motion Builder before plotting it to a CG actor. Once the actor had movement the next step was to set the character up in 3D Studio Max. The original bone system was kept and the character was exported as an FBX file recognisable by Motion Builder. The file was imported into the same scene as the actor with the motion plotted to it and an automatic rig was created for the character. At this point the character is ready for either animating or adding mocap data. A final step was made to change the character input allowing the character to follow the same movement as the actor.

Research Artefact 4: Mocap Tests with Suit

This artefact tests the motion capture suit made in the previous artefact and shows the benefits of its use through different movements. Movements decided on were chosen for the ease in which the cameras can process the movement. Camera positioning was altered for this capture session to allow a central position for the actor to limit the amount of undesired reflections. Initial tests were made with simple arm and leg movement, which the cameras picked up with improved accuracy over the previous lab sessions without the use of a suit. Still some problems remain as other reflections were still being picked up and were corrupting the correct data. Another test with the same movement was performed but now with lower lighting and exterior reflective objects covered. This showed an improvement in the results with less reflections being picked up and improved stability of the reflective markers. Another capture was performed using a simple walk to see whether this can be performed with the same accuracy. Results showed limited reflections but some cleaning of the data was required as elbow markers got confused with hip marker due to their close proximity. The next capture performed was to analyse whether the system could now handle rapid movements produced from fast punches and kicks. Results showed a disappointing capture as rapid changes in direction resulted in prolonged marker disappearance. This caused the automatic cleaning process to fill in the missing markers with undesired translations and caused deformed movement. A solution to this problem can be found when the motion is plotted to a character, as then the motion is converted to key-frames and can be animated.

Research artefact 3: Making Mocap Suit

This artefact explains the process for creating a motion capture suit and how it may improve the motion capture process. A motion capture suit is tried out due to multiple problems trying to a get a clean capture in the previous artefacts. First problem encountered when trying to get a clean capture was that the reflective markers had trouble sticking to the actor’s skin and during rapid movement would fall off. This was solved by attaching the markers to the suit using Velcro and fabric glue. The second problem was with reflective markers disappearing throughout the capture due to the back ground being too bright. The third problem was with the markers getting confused with each other for two possible reasons, the first being the lightness of the background and the second being the close proximity between them. A motion capture suite was developed using black sport under armour, which is a skin tight top and trouser set made of elasticised fabric. This was to create a dark surface to attach the reflective markers to and helps the cameras detect them. The choice of this clothing was based on research on existing motion capture sessions performed to create motion inside video games and films such as resident evil 5 and for Golum in The Lord of the Rings. The decision on the amount of reflective markers to use and their placement on the suit was based on the MotionBuilder marker sets for the CG actor and successful captures performed by others and published on the internet. The suit was created using the minimal amount of markers to avoid too much confusion for the cameras, as well as to cut down on the amount of time spent cleaning the data later on.

Research Artefact 2: Plotting Mocap to 3D Actor

This artefact demonstrates the difficulties associated with capturing motion and plotting it to a computer generated character. The first motion capture lab session was ineffective at producing the correct amount of dada needed for creating a character rig for the CG actor. This data when plotted to the actor produced a frozen state where none of the body parts had motion. A second lab session was taken with corrections made based on the analysis of the first motion capture lab session. More reflective markers were used in this session raising the amount used from 12 in the first session to 48 in the second. This larger amount of markers aloud a successful character rig to be created for the CG actor. This however now raised a new issue of marker detection being difficult, as throughout the capture markers were getting confused with each other due to the short distance between them. Major areas affected by this were the hands, wrists and elbows and produced erratic behaviour when the data was cleaned up. The motion capture software used utilises an automatic data cleaning system, which fills in the gaps for missing markers by predicting intended trajectory based on the existing motion curve. Gaps occur when marker area blocked by extremities or due to having to much light in the room or on the surface the marker is attached. When large gaps appeared in the data produced in the second lab session the automatic cleaning system filled in these areas leaving the motion deformed. This was also aided by individual markers disappearing and the cleaning system thinking a nearby marker was the initial marker. This produced negative affects and deformed the motion further.

Personal Research Artefact 2: 2nd Mocap Lab Session

A second motion capture lab session was arranged for the second artefact created for the research project to correct the problems of the previous session. After cleaning up the data for the first mocap session, it was plotted to a 3D actor within MotionBuilder. This however had unexpected results due to the not having the required amount of marker point to drive the character. After testing the marker set up it was apparent that head, feet and limb markers needed to be added to drive those areas, as the lack of these resulted in the character freezing up. In the second mocap lab session more marker points were added to correct this problem. This resulted in a much more effective result for MotionBuilder to recognize the movement. This capture brought to light another issue in the form of how effective the cameras were in detecting the reflective markers and keeping them locked on a point. The cameras struggled to monitor these markers because of reflective areas in the room and on the clothing. Also each marker needed to be in the view of at least three cameras at a time for an effective capture and this was difficult to achieve with a basic 7 camera set up. This capture then had floors in marker detection and resulted in large gaps where markers disappeared or got confused with a different marker on the body which pulled the wrong extremity to desired location. The clean up for the disappearing markers had undesirable effects and brought body parts away from the correct area and resulted in deformed motion. Also noticed from the final plotting of the motion data was that different movements resulted in either a successful or unsuccessful capture, slower dancing was more effective then faster boxing.

Personal Research Artefact 1: Motion Capture Lab Session

The motion capture lab session takes advantage of a motion capture facility which was originally built for the analysis of movement for biomechanical research within sports. The facility has the ability to capture motion for export in a c3d file format, which is Autodesk MotionBuilder friendly. A semi pro Thai Boxer is used as an actor to look at how the hardware reacts to specific martial art motions. The sets of motions are specifically chosen to look at a range of issues that need to be considered when animating. These consist of mass and weight, speed, sudden stops, spins and follow through. These issues are focused on to establish whether an individual element within motion creates more problems than another when the process of motion capture is in effect. The other area of interest is whether or not this method for creating motion within CGI is efficient for the motion needed. This area consists of the time taken to organise the motion capture lab session and to capture the desired motion while also taking into consideration any cleaning up of the data needed. Also any extra costs associated with the overall process.

Fast and spinning motions performed by the actor, such as the fast kicks and the spinning kicks, created difficulties within the motion capture. This was due to the cameras struggling to keep up with the refractive markers attached to the actor, as well as certain markers being obstructed from view by body parts. This created areas of the motion capture data to have blank spots, this intern added time to the cleaning up of the data as well as creating inaccurate motion.

Research Document Artefact Breakdown

Areas Studied:

Motion
• Mass/weight (jumping and landing)
• Speed (continuous fast movements)
• Sudden stops (punching and kicking pads)
• Spins (various combinations involving spins)
• (Optional) Interaction with another person (sparing)

Software
• What preparations need to be made?
• How does the software work?
• What are the problems and difficulties that are associated with it?
• How does it react to different motions?
• How long does the process take?
• How much cleaning up of the data is involved after data collection?
• How is the software usually used being that it’s part of the Sport Science faculty?
• Based on the research paper, are there any other methods that can be used at this facility to aid with character creation e.g. capturing skin deformation and muscle movement. Based on this answer, alternative artefacts may be made.

Key-Framing
• Replicate the motion capture movement through key-framing a character.
• Record how long the process takes.
• Record the different processes taken.
• Note any difficulties involved.
• Use the video reference to aid the key-framing.


Artefacts:

Artefact 1: Motion Capture

Motion studied: Mass/weight.
Action: Flying knee, flying elbow, jumping round kick.

Details:
How does the software react to the motions? This experiment shows how the data recorded can be used to give mass and weight to a character. The motion will be mapped to a character of the same shape and size as the actor as well as to a character of a different shape and size.

Artefact 2: Key-Framing

Motion studied: Mass/weight
Action: Flying knee, flying elbow, jumping round kick.

Details:
The character of the same shape and size of the actor will be key-framed to do the same movements. This shows how mass and weight are produced in animation. Similarities and differences between the mocap technique and the key-framing will be noted.

Artefact 3: Motion Capture

Motion studied: Speed.
Action: Multiple punch and kick combinations on pads.

Details:
How does the software react to this motion? This experiment shows how effective motion capture is in recording fast movement. The motion data will be mapped to the character of the same size and weight of the actor.

Artefact 4: Key-Framing

Motion studied: Speed
Action: Multiple punch and kick combinations on pads.

Details:
The character of the same shape and size of the actor will be key-framed to do the same movements. This shows how fast movements are produced in animation. Similarities and differences between the mocap technique and the key-framing will be noted.

Artefact 5: Motion Capture

Motion studied: Spins
Action: Round kick, spinning elbow/back fist, back kick, spinning kick.

Details:
How does the software react to this motion? This experiment shows how effective motion capture is in recording movements which spin the reflective markers attached to the actor. Will there be need for much time consuming clean up of the data. The motion data will be mapped to the character of the same size and weight of the actor.

Artefact 6: Key-Framing

Motion studied: Spins
Action: Round kick, spinning elbow/back fist, back kick, spinning kick.

Details:
The character of the same shape and size of the actor will be key-framed to do the same movements. This shows how spinning movements are produced in animation. Similarities and differences between the mocap technique and the key-framing will be noted.