Feasibility of markerless motion capture: achieving complex, multiactor driven character motion using lowend equipment and software
The process of this project demonstrates the feasibility of capturing complicated movements, through markerless motion capture using lowend technology, for freelancers and small studios.
Due to the high cost of the modern and fast working technology, it is not easy for small studios and freelancers to obtain similar outstanding outcomes. The research conducted through the honours project, demonstrates the possibility of adding advanced markerless motion capture data onto a rig, using lowend softwares and hardwares, by using lowend equipment.
In 2015 there was several motion capture tutorials available for the public which were using a single Kinect device, tracking a single person. There was some additional sources for multi person detection with a dual-sensor setup as well, however, these setups were really limited.
As a baseline for this project, the goal was to track complex motion where the actors are interacting with each other.
Further more, the project meant to demonstrate if it is possible to record a scene in a significantly smaller area, than what is shown on screen.
The entire dance was recorded using 3 Microsoft Kinect in a 6 by 5 ft. room. The 3 devices made it possible to track multiple actors with fine movements.
Through tests, the aim was to demonstrate the ability to generate an animation with complex human motion using markerless motion capture equipment and software. The human motion data was transferred onto two rigs, demonstrating the possibility of generating advanced and complex motion through dance, with the addition of some special effects.
Several levels of error correction and fine tuning was applied to the original motion data, before merged to skeleton and rigs.
After the motion snippets were imported to Maya, further clean up and editing of the scenes was required. Each scene had to be well positioned, both in time and in the 3 dimensional World.
Additional special effects like looping a take, to create a long walk, was applied after.
After the editing, further effects such as the smoke disassembling and textures were added.
The original plan for the project contained several different types of work, which were all new areas with no prior knowledge. These specific areas such as motion capturing, cleanup, scene editing, 3D, special effects and fluid simulation are complex fields requiring great knowledge and experience. With an eager plan, these areas were underestimated by time and knowledge. Should a similar markerless motion capture project appear, comparing the required work and achievable outcome to the Kinect 2 and PS Eye devices, could prove interesting. Furthermore, a setup with better lighting, more recording devices and a bigger volume may reduce the tracking errors dramatically.
Adobe Premiere Pro