
Previous Post
1.WHY CREATING DIGITAL HUMANS IS IMPORTANT?
High-fidelity digital humans must be created for a variety of uses, such as immersive
telepresence, AR/VR, 3D graphics, and the emerging metaverse. The process of creating
personalised avatars is time-consuming, expensive, and usually necessitates the calibration of
multiple cameras.
2.WHICH SPEEDING UP THE RENDERING PROCESS?
To establish the colour of a pixel in standard volume rendering, it is necessary to query and
compile densities at hundreds of points along the ray. One common method to expedite this process
is to keep an occupancy grid that skips samples in the empty space.
When given the same time budget as SoTA approaches, their technique performs noticeably better.
They also offer an ablation study to demonstrate how the components of their system effect speed
and accuracy. On GitHub, the code will soon be available.
3.HOW MONOCULAR NEURAL AVATAR IS HELPFUL FOR INSTANTAVATAR?
By outlining a method that takes just as long to acquire input video as it does to rebuild it,
they expect to significantly advance the deployment of monocular neural avatar reconstruction in
practise. A system that can recreate high-fidelity avatars in 60 seconds as opposed to hours, given a
monocular video, posture settings, and masks.