LogIn
The team proposed a capture pipeline that can translate a variety of different facial hair styles and lengths into 3D.
A team of scientists from Disney Research Studios published a new report on facial hair tracking for high-fidelity performance capture, a new method of capturing various facial hair styles and skin underneath it. In this report, the team proposed the first multiview reconstruction pipeline capable of tracking 3D facial hair, ranging from sparse and short to dense beards, as well as the underlying 3D skin for entire performances.
The team described the method of creating a hair-free reference mesh for tracking, spoke about tracking the reference mesh with the facial hair mask, explained the process of tracking the reference facial hairstyle, showed how the mesh is refined, and shared tons of formulas that thoroughly explain the algorithm behind the proposed method.
"In this work, we tackle the complex problem of facial hair tracking for high-fidelity performance capture. Our method is able to reconstruct and track individual facial hairs over complex performance sequences in a traditional multiview reconstruction scenario. We additionally create a realistic approximation of the dynamic cleanshaven facial surface, as if the actor had been captured without facial hair, thus removing the need to actually shave," commented the team.
You can learn more and read the full report here.